MIDWAY REPORT
ANTACCS PROJECT
July 1964
INFORMATICS INC.
MIDWAY REPORT
ANTACCS PROJECT
TECHNOLOGY, METHODOLOGY AND INTEGRATION
COVERING PERIOD
January 1964 to July 1964
Prepared Under Contract
to
The Office Of Naval Research
(Nonr-4388(00))
by
INFORMATICS INC.
15300 Ventura Boulevard
Sherman Oaks, California
July 15, 1964
ERRATUM
Discontinuities in the numbering system omit page numbers
2-166 to 2-177
2-245 to 2-265
4-28 to 4-33
No material is missing.
TABLE OF CONTENTS
Page No
i
i i i
TABLE OF CONTENTS
LIST OF ILLUSTRATIONS
SUMMARY V
SECTION I INTRODUCTIONS
1-1 General 1-1
1-2 Study Objectives and Approach 1-4
1-3 Report Organization 1-5
SECTION 2 TECHNOLOGY
2-1 General and Introductory 2-1
2-2 Display Technology 2-2
2-3 Displays - User Technology and Software 2-23
2-k Input/Output Technology 2-77
2-5 Memories 2-131
2-6 Components and Packaging 2-189
2-7 Advanced Usage Techniques 2-217
2-8 Computer System Organization 2-221
2-9 Programming 2-268
SECTION 3 METHODOLOGY
3-1 Introduction 3-1
3-2 General Methodology 3-2
3-3 Implementation Methodology 3-82
3-4 Specific Methodology 3-113
SECTION 4 STUDY INTEGRATION TASK
4-1 Scope and Objectives of Study Integration Task 4-1
4-2 Comparison of Implications of Alternate System
Operating Concepts 4-6
I I
SECTION k (Cont'd.)
Page No
k-3 Demonstration of the Synthesis and 4-21
Evaluation of a System Node
k-k Discussion of System Planning I terns k-3^
SECTION 5 BIBLIOGRAPHY
5-1 Introduction 5-1
5-2 Technology 5-2
5-3 Methodology 5-^7
I I I
LIST OF ILLUSTRATIONS
FIG. NO. TITLE PAGE NO
2-1 Man/Machine Coordination 2-31
2-2 Over-all Command Function 2-33
2-3 List Display of Military Installations 2-35
2-4 List Display Modifying Military Installations 2-35
2-5 List Display Modifying Bomber Air Bases 2-35
2-6 List Display Modifying Fuel Storage 2-35
2-7 List Display Political Limits 2-36
2-8 Format Display Current Totals 2-36
2-9 Schematic of Commander's D= P. System 2-40
2-10 Typical Operator Steps in Use of Function Keys 2-44
2-11 Computer Steps in Conjunction with Function Keys 2-45
2-12 Computer/On-line Device Configurations 2-46
2-13 Example of Format Display 2-50
2-14 Selection "Trees" 2-51
2-15 Console/Processor System Operation 2-53
2-16 Basic Executive Control Loop 2-61
2-17 Tasks Associated with Scanning Input Message Lines 2-62
2-18 Processing Servicing Requirements 2-64
2-19 Programs for a Display Console 2-73
2-20 Probability of Console Service 2-75
2-21 Relations of Man and Machine 2-78
2-22 Typical Display Overlay 2-93
2-23 Typical Series of Operator Steps 2-95
2-24 Rotating Drum Printer 2-98
2-25 Impact Wheel Printer 2-99
2-26 Matrix Printer 2-100
2-27 Stylus Printer 2-101
2-28 Chain Printer 2-102
2-29 Stick-Type Printer Bars 2-103a
IV
FIG. NO. PAGE NO
2-30 Electro-Optical Printer 2-108
2-31 El ectrograph i c Printer 2-110
2-32 Magnetic Printer 2-113
2-33 Drum Printer 2-1 16
2-3^ Typical Interface Functions 2-122
2-35 Storage Capacity and Cycle Time of Memories 2-182
3-1 Command and Control Environment 3-15
3-2 Functional Diagram of Air Traffic Control
Simulation 3-16
3-3 Hypothetical System Design 3-29
3-^ Missile Interceptor Model 3-32
3-5 Design Optimization Problem 3-^6
3-6 Carrier Transmission System 3-^7
3-7 Technical Branches PMR 3-5^
3-8 Manual Vectoring Schematic 3-62
3-9 F-4b Intercept Simulator 3-67
3-10 ATDS Prime Avionics Equipment Configuration 3-71
3-11 Radar Data Input Simulation 3-78
3-12 Using the Cockpit Simulator 3-78
3-13 Communication Subsystem Simulation 3-81
3-1^ Communications ^ 3-81
3-15 System Definition Phase 3-9^
3-16 System Design Phase 3-95
3-17 Program Design Phase (l) 3-96
3-18 Program Design Phase (ll) 3-97
3-19 Program Production Phase 3-100
3-20 Program Test Phase 3-101
3-21 System Test Phase 3-102
3-22 System Operation Phase 3-103
ANTACCS MIDWAY REPORT
SUMMARY
This is a report of the technology, methodology,
and integration aspects of the ANTACCS study project
sponsored by the Office of Naval Research, in conjunction
with various Naval organizations including BuShips,
BuWeps, Chief of Naval Operations, and Commandant Marine
Corps. It is the preliminary report of work performed by
Informatics Inc. under Contract Nonr-4388(00) . it is a
Midway Report representing that portion of the work
accomplished during the first half of the 12 month
project.
The project members of Informatics Inc. are in-
debted to Mr. R. Tuttle, the ONR Scientific Officer who
is guiding this effort. They are also indebted to the
Study Monitor Group, a group consisting of knowledgable
and experienced persons from BuShips, BuWeps, Chief of
Naval Operations, NAVCOSSACT and Marine Corps Headquarters
for their advice in this effort. The knowledge of the
Scientific Officer and the Study Monitor Group in Naval
requirements and environment as well as their experience
with present Naval efforts has been of valuable assistance
in the assurance of a more useful product.
The technical staff of Informatics has been
supplemented in certain technical areas by subcontracting
efforts of Hobbs Associates. Hobbs Associates has
provided many of the sections on hardware techniques,
especially those in the circuits and packaging areas.
Hobbs Associates has also contributed in the area of
memories and display devices.
I
I
VI
The purpose of this project is to develop and
present Information concerning technology, methodology
and integration which will be of assistance to planners
in the design and implementation of command control
systems. The project scope and emphasis is restricted
to the application of its techniques and data to the
solution of problems concerning the Advanced Naval
Tactical Command Control System. This system is identi-
fied by an SOR as being visualized for the 1970-1980
time period, and for which a TDP is to be developed in
1956.
The three areas of the project are: Technology,
Methodology, and Integration. Technology deals with
scientific and technical material of potential use to
Naval command control systems. This material Includes
both hardware and software subjects. Methodology is
concerned with technical and managerial techniques used
in the planning and implementation of Naval systems.
I ntegratlon covers the unification of technology,
methodology, and requirements into candidate approaches
to the design of Naval systems and their parts.
This Midway Report represents work In progress.
An attempt has been made to organize the report in
such a way that all areas are included, at least structurally,
as they will appear in the final report. Since this is a
preliminary report, many sections are incomplete. In many
cases only a foundation of information for the technical
area has been collected and organized. Still to be accom-
plished in most of the areas is the translation of that
basic information to opinions and conclusions concerning the
usefulness or future application of this particular information
For instance. In the display area the techniques have been
iden.tlfied, classified and analyzed. Remaining Is a discussion
VI 1
of the relative merits of the various approaches,
especially related to operational requirements in
ANTACCS.
it Is the express desire of the Informatics
project personnel that this report be examined with a
critical eye by the Scientific Officer, the Study Monitor
Group and others who may also be qualified to criticize
it. It is hoped that the critical viewpoint is taken
within the framework of the fact that this is a pre-
liminary report representing work in progress. The
major useful contribution of the contract effort will
be developed and reported during the last half of the
project effort. We believe that sufficient Information
Is presented In this present report to Indicate the
general direction In which the project is proceeding.
It should provide the basis to allow the ONR Scientific
Officer and the Study Monitor Group to redirect the
efforts as appropriate.
1-1
1. INTRODUCTION
1.1 GENERAL
Perhaps the fastest growing technology within the military at the
present time is that of command and control systems for strategic and
tactical uses. This technology deals with the application of modern
electronic computer techniques to modern military operational requirements.
The subject matter of this report is techniques for design and implementa-
tion of tactical command and control systems. Although the work deals
principally with the techniques, equipments, technology and methodology
involved in the implementation of Advanced Naval Tactical Command and
Control Systems (ANTACCS) for the 1970-1980 time period; the information is
of interest in other military command and control systems.
The technology stemming from and involved with the modern stored
program electronic computer is only 14 or 15 years old. It has not
benefited from the many years of experience and discipline of such Impor-
tant Naval technologies as naval architecture or armament design. The
technology is very new and rapidly growing; methodology in the sense of
unified, universally-used techniques is virtually nonexistent. The object
of the ANTACCS study project is to provide information on information
processing technology and information systemsdesign methodology to serve as a
resource document for the use of planners of future Naval tactical data
systems.
Military command and control systems provide a greater challenge to
electronic data processing than any other application. Some of the
characteristics of command and control systems which account for this
chal lenge are:
1) Considerable emphasis on man/machine interactions and
requirements; the command and control system is, in
the final analysis, designed to facilitate the decision
processes of the commander and his staff.
1-2
2) The computer is imbedded in an on-line or real-time
environment involving a large amount of instrumentation
and peripheral equipment with which it must interact.
3) In most command and control applications there are large-scale
file management tasks involving the receipt, collation, and
retrieval of facts.
4) Most command and control system are very large information
handling systems involving a multiplicity of input and output
information channels. These systems are further complicated
by the requirement to interact logically with many command
levels and with much remote instrumentation.
5) There are especially challenging operational and environ-
mental problems; reliability requirements are exceedingly
high, and physical and logical environmental constraints
are very often especially restricting.
Together these create a special challenge for the designer, from the
component level to the system integration level.
Naval tactical command and control systems supp ly even greater
challenges than command and control systems in general. There are
additional physical factors which become important; space limitations
and the movement of a ship or vehicle are simple examples. There are
additional logical and operational factors: missions change for a
given vehicle or platform, the entities with which the system communicates
may change depending on the mission or operation, and there are extra-
ordinary problems of logistics in installing, maintaining and operating
equipment on a fighting ship.
To assure a meaningful product and remain responsive to the requirements
of ANTACCS, the project will illustrate the techniques developed by applying
them to the requirements developed and discussed by Booz Allen Applied
Research, Inc. in a companion report. Accordingly, this volume discusses
1-3
technical material useful to system designers in defining, designing,
and implementing systems such as ANTACCS. This is done from the
standpoint of the electronic equipments involved, and with special
regard for the people who will use them. It deals with the technology
of future command and control systems, that is, the hardware and
software techniques available for or necessary to system implementation
It also deals with the methodology of system implementation, that is,
the techniques of system engineering management and the application
of these techniques to satisfy requirements and to thereby produce
an operational system. Although this work Is related to the require-
ments which have been developed in a companion project by Booz Allen
Applied Research, Inc., to a very great extent, it can stand on its own
as a document for future use of planners of Naval command and control
systems.
The scope of the present work is of extraordinary magnitude. The
subject matter of technology ranges from integrated circuitry of a
computer to computer systems organization. Methodology subjects range
from simulation languages to techniques for planning and implementing
ANTACCS.
1-4
1.2 STUDY OBJECTIVES AND APPROACH
There are three aspects of the study treated in this report:
Technology, Methodology and Integration. Technology deals with the
techniques and embodiments - both hardware and sof tware--for implementing
data processing functions. Methodology deals with the techniques for
the design, evaluation, and synthesis of equipments of all levels within
the system as well as the management techniques for accomplishment of an
operational system.
The principal objective of the integration subtask is to illustrate
technology and methodology by developing certain approaches to the design
of various portions of ANTACCS, as prescribed by the requirements developed
by BAARINC. It is visualized that this subtask will analyze and evaluate
synthesized system components at various levels, thus illustrating how the
various aspects of technology and methodology are unified and integrated
into usable concepts.
The objectives of the ANTACCS project are visualized to the the following:
1) To identify, analyze and evaluate hardware and software
techniques of potential use in ANTACCS.
2) To develop resource information and to provide reference
documentation representing information of use to future
Naval command and control system planners.
3) To supply a unifying force to integrate the concepts
developed or available in technology and methodology.
4) To develop approaches to a number of candidate systems
which illustrate, in a practical way, techniques of
technology and methodology in ANTACCS.
The information developed in this project can be used in a number of
ways:
1) As reference documentation.
2) To identify research and development needs for future system
implementation.
3) As a-specific guide to planners of future systems.
1-5
1 c3 REPORT ORGANIZATION
The principal organization of this report is a division of the
presentation into the three main efforts: technology, methodology
and integrationo Section 2 covers technology. The following items
are covered: displays, input/output devices, memories, components
and packaging, packaging techniques, advanced usage techniques and
machine system organization.
In each portion of the technology there is first a classification
of the techniques. Next, the sources of information are discussed
and presented. This refers to the people, companies, and the
literature from which information was obtained. The characteristics
appropriate for ANTACCS are discussed, the application of the
technology in the Naval environment is further presented, as well
as a review of the current status of the equipments and techniques.
Following this, the availability of technology in the 1975 &ra^ the
limitations of the present and planned technology and the recommended
developments for the future are presented. Each portion is concluded
with a discussion of evaluation criteria, conclusions and recommendations
Although the specific sections may deviate from this order in certain
instances, in general in each section there is an attempt to cover
all of these poi nts .
Section 3 deals with methodology. For the purposes of project
organization, methodology has been split into three major areas:
general methodology, implementation procedures, and specific method-
ology. General methodology deals with:
1) Tools and techniques which the system designer has at
his d i sposal .
2) A methodology generally or universally applicable and not
necessarily restricted to systems of a special purpose
nature or a special class.
3) A methodology which is generally available, exists as a
tool, and can be readily applied.
-6
Implementation procedure deals with the understanding of tasks - both
technical and managerial - which must be accomplished in the implementa-
tion of modern command and control systems. Specific methodology
relates to the special requirements of selected equipment configurations
and design problems which might arise in ANTACC5.
In general, the elements discussed above under technology with
regard to classification, sources, requirements, status, limitations,
evaluations, developments, conclusions and recommendations are
covered in each methodology section. Under general methodology
simulation languages and techniques of simulation are covered in
some detail. Under implementation procedures, system design, implementa-
tion and evolutionary aspects of systems have been described. In
specific methodology some quantitative design tools are presented.
In the section on integration the scope and objectives of
integration are first discussed. Following this there is a comparison
of implications of alternate system operating concepts. This concerns
the various operational philosophies which the Navy might adopt relating
to the structure and organization of the various tasks to be performed
and involving such aspects as platforms, missions and command structures,
Following this there is a discussion of the synthesis and evaluation
of a system node. This illustrates how the information developed under
requirements with regard to system technical functions can be trans-
lated into data processing functions. A matrix technique is described
which relates platforms together with their missions and command
levels, to operational tasks and data processing tasks.
Included in each section is an extensive bibliography. Generally,
in the text reference is made at appropriate points to items in the
bibliography. Also, at the close there is a general bibliography for
the entire project effort.
(in the future final report there will be certain additions and
modifications to the organization described above. For instance,
in the final report there will be an extensive list of opinions,
conclusions and recommendations for the entire project effort and
for the various tasks. There will also be a cross reference index
to assist the reader in finding his way through the technical
information. Another item to be added in the final version is an
extensive glossary of terms.)
2-1
2. TECHNOLOGY
2>1 GENERAL AND INTRODUCTORY
The purpose of this study of present and advanced technology is
to identify, analyze, evaluate and document those areas of technology
which will have significant impact on future Navy tactical data
systems. This study will probably not uncover any new areas of
technology or disclose any new areas of application to the experienced
system designer. It will, however, provide analyses which will place
the new technologies in proper perspective and provide criteria and
examples to aid in evaluating and selecting future equipment.
As the first phase of this study is necessarily devoted to the
collection of information and the analysis of that information, rather
than its evaluation and documentation, it follows that little completed
work is available to include in this report.
It is, therefore, intended that this section should indicate those
areas in which progress has been made and show the type of work being
done, rather than present some small sample of the finished product.
2-2
2.2 DISPLAY TECHNOLOGY
2.2.1 Classification of Display Types
Display types can be classified in a number of different ways
that are not mutually exclusive. Associated groupings of display
technology will vary with the method of classification. Among the
ways in which displays can be classified are:
1) Functional Console
Large-Screen
2) Nature of Data to Status Displays
be presented Real-time or Dynamic Displays
3) Type of Data Alphanumeric
Symbol s
Graph ical
k) Type of Mechanization Cathode ray tube
El ectrol umi ne scent
Character 1 i ghts
Photographic Projection
L i ght-val ve
Mechanical Inscriber
Photochromic
Ferro-el ectr i c
Opto-Magnet ic
Laser-1 umi nascent
In this report, displays will be classified by the type of mechani-
zation. In the discussion of each type of display that has been investigated
to date, the functional use and the nature and type of data to which it
is adaptable will be considered. Any factors that make a particular
type of display unsuitable for a certain function or for certain kinds
of data will also be noted. For example, the fact that a cathode ray
tube is not suitable for a large-screen display, or the fact that a
photographic projection type, large-screen, display cannot present
real-time dynamic information, will be discussed as limitations of
these techniques that make them unsuitable for certain functional uses
and for certain types of data.
2-3
2-2.2 Sources of Information
2.2.2.1 People and Organizations
The following lists the companies and governmental agencies with
whom displays have been discussed during this study, and the type of
displays discussed with each.
1) Bunker-Ramo Corporation Light-valve displays
Canoqa Park, California r- • • ■ . l-
^ Continuous-strip photographic
projection displays
CRT d i spl ays
2) General Dynamics/Electronics Charactron CRT displays
San Dieqo, California , . .^ i j- i
^ Light-valve displays
3) General Telephone Laboratories Continuous-sheet electroluminescent
Bayside, Long Island, N. Y. displays with XY matrix addressing
Acoust ic/el ect rol umi ne scent
d i spl ays
4) RCA Laboratories Ferroelectric displays
Princeton, New Jersey
5) Laboratory for Electronics Magnetic thin-film displays
Boston, Mass.
6) NCR Electro-mechanical photochromic
El Segundo, California displays
CRT-Photochromi c projection
d i splays
7) Sylvania Discrete alphanumeric character
Waltham, Mass. displays
Continuous-sheet electro-
luminescent displays with XY
matrix selection
8) Stanford Research Institute Magnetic thin-film displays
Menlo Park, California Modulated crystal filter
d i splays
2-k
9) Rome Air Development Center
Rome, New York
Light-valve displays
Modulated crystal interference-
f i 1 ter d i spl ay s
Thermo-p 1 ast i c displays
Laser d i spl ays
10) USAER&DL
Fort Monmouth, New Jersey
11) U.S. Navy Bureau of Ships
Washington, D.C.
Photochromic displays
Laser-luminescent displays
Fiber-optic CRT displays
Photographic projection displays
Di spl ay memor i es
Photographic projection displays
CRT di splays
Much of the information presented in subsequent portions of this
section are based on discussions with display experts in the organizations
listed above. Their descriptions of specific display techniques and
their opinions of the advantages, disadvantages, and limitations of
display techniques were relied upon heavily in the preparation of this
report .
2.2.2.2 Literature
An extensive list of references pertinent to the study of display
technology are given in the Bibliography. To date, only a few of these
have been studied in detail. Some of the material in this section has
been extracted from these references. The more important and pertinent
of these references will be studied in detail during the remainder of
this study and new references will be added to the Bibliography to
r-eflect material published or discovered subsequent to the preparation
of this report.
2.2.3 Display Characteristics for ANTACCS
The display characteristics required for ANTACCS cannot be fully
identified at this time since the results of the requirements analysis
have not been available. However, it is anticipated that both console
and large-screen displays will be required; that alphanumeric, graphical,
and dynamic real-time data must be presented; and that multi-color
displays (particularly for large-screen applications) will be required.
2-5
It Is further believed that electro-mechanical display systems and
photographic projection systems will not be acceptable for a 19/0 system
The analysis of display technology and the information presented
in the final report will permit the selection of display technologies
with the appropriate characteristics for any functional use in an
NTDS or MTDS system. This analysis will include all the more important
and feasible types of displays that might be applicable to such systems
in 1970.
2.2.4 Applications of Displays in the Naval Environment
It is anticipated that applications of displays for shipboard and
ground-based military environments in the 1970 era will include console
and large-screen presentation of both status information and real-time
dynamic Information such as target track data. The applicability of
specific types of displays to different applications will be considered
in further detail in the remainder of this study.
2.2.5 Current Status Review
The investigation of new display technology has not progressed as
far at this point In the study as that of some other areas, such as,
for example, memory technology. The Information collected to date Is
summarized here, but comparisons and evaluations of different types
of displays are not yet available. Detailed comparisons of specific
existing display devices were presented and discussed In the Initial
proposal for this studyc Although these are available, they are not
included in this report since they are representative of past technology
rather than that to be anticipated for the 1970 era. Since a number of
satisfactory techniques for console type displays are now available,
there will be no problem with respect to the availability of console
type displays for a 1970 system. Existing cathode ray tube technology
and anticipated improvements in this technology should meet all re-
quirements for small-screen console type jJisplays, even if none of the
new technologies prove to be superior. However, with respect to
- References are listed at the end of each subsection.
2-6
large-screen displays the situation is much less favorable. In an RADC
Technical Documentary Report published in 1962, the state-of-the-art
and development efforts for large screen displays were described as
fol lows :
"Display developments are being undertaken in three major
technological areas. These areas may be differentiated in
terms of the basic processes being applied and on the basis
of development time required to provide fully operational
subsystems .
The first of these processes is based on projection and employs
a stable light modulator, such as film or selenium plate, to
provide the display. Operational subsystems of this sort are
considered to be achievable within months.
The second process, the light valve, in theory should provide
adequate performance for systems applications, and it has the
dual advantages of operation at electronic speeds and of the
elimination of expensive film. However, the performance potentials
have not been realized in practice, and major technological
improvements must be made before the light valve can be useful
for most systems applications. The presently available models
exhibit major weaknesses in their capability to provide high
resolution and brightness.
This low brightness makes it impossible to use the light valve
in the high-ambient lighting conditions of most of the systems.
The interactions of the oil film and the lens systems are such
that it is not possible to increase the display brightness
level without major improvements in the characteristics of the
modulation surface. Improvement, very likely, is contingent
on the development of suitable thermoplastic materials. Light
valve techniques show considerable promise, and with suitable
development may eventually supersede film systems. However,
it should be clearly recognized that full realization of the
light valve's potential may require years of additional research.
The third process, electroluminescence, does not require projection
since the display surface itself acts both as light source and
modulator. Only small laboratory devices for demonstration and
experimentation are available at the present time. Electroluminescence
is appealing in Its apparent simplicity, its capability to eliminate
projection, and Its characteristic of non-catastrophic failure.
In addition, there Is a potential for full color operation at
high brightness levels, and the large surface reduces the problems
of obtaining high resolution. Unfortunately, there Is an Impressive
number of technical obstacles that must be overcome before electro-
luminescent devices can meet the requirements of the systems. The
2-7
most immediate problem is that of modulating the display surface,
and a number of promising efforts are underway in this area at
the present time. This effort is concurrent with others that
are aimed at the development and application of new phosphors
to obtain high brightness levels and multiple colors. However,
even allowing for impressive technological improvements, years
will be required to advance the capability of electroluminescent
displays to the point where they can serve as dynamic large scale
displays for system applications.
Desirable as these advanced displays are, most immediate require-
ments of Command and Control Systems can only be met by projection
techniques using film or xerographic techniques for light modu-
lation."
Unfortunately, developments during the past two years have not
significantly altered the views quoted above, except that improved
technologies are of course somewhat closer to realization now than they
were in 1962.
Photographic projection techniques are still the only feasible
means of meeting operational requirements for large-screen displays in
Command and Control Systems. Significant progress has been made in
light-valve type displays during the last two years, but the reliability
and life of these devices does not permit their use at this time in an
operational system in which minimum down time is an important require-
ment. However, new and improved light-valve type devices offer great
promise for a system to be operational in 1970.
Display techniques that have been developed or that appear
promising for the future include individual character lights, cathode
ray tubes, mechanical inscriber systems, film or photographic projection
systems, light-valves, photochromic systems, electroluminescent devices,
ferroelectric devices, opto-magnet ic devices, and laser systems. Of
the above techniques, it is believed that mechanical inscriber systems
and film and photographic projection systems will be obsolete by 1970.
Improved light-valves, electroluminescent panels, photochromic displays
and, possibly, laserl umi nescent displays appear very promising for that
time period. The display technologies that have been investigated to
date are discussed briefly in the following parts of this section.
2-\
2.2.5.1 Mechanical Inscribing Machine
A mechanical inscribing system permits the large-screen display
k
of real-time dynamic information at a relatively slow rate . In this
type of display, a glass slide coated with an opaque material is in-
serted into a projection system. Another glass plate with a stylus
mounted in its center Is positioned parallel to the first slide so that
tipping the glass plate causes the stylus to penetrate the opaque material
When the stylus is moved In the X and Y directions by a servo-mechanism
under the control of external signals, a trace is inscribed in the
opaque material on the face of the slide. The light from a lamp is
projected through this trace on the glass slide and focused on a
projection screen. Thus, the trace will appear on the screen and can
be drawn In real-time.
The use of color filters in the light path permits color traces
to be generated. A composite multi-Input or multi-color display can
be generated by superimposing the Images from several projection systems.
Additional projectors can be used to superimpose static information,
such as maps, on the dynamic Information. Since the inscribed trace
remains on the glass slide, no external memory Is required for this
type of d I spl ay .
With a trace width of 0.001 Inches on the slide, the projected
trace will be about 0.1% of the screen size. Recent systems require
approximately 50 milliseconds to Inscribe a trace across the full width
of the screen. Alphanumeric characters can be Inscribed at a rate of
approximately 20 characters per second.
2.2.5.2 Charactron Shaped-Beam Cathode Ray Tube
The Charactron Is basically a cathode ray tube which includes
a character generating mask and the necessary electrodes for shaping
the beam Inside the tube « The electron beam is deflected to the
proper position in the character mask corresponding to the character
to be generated. As the beam passes through the mask, It is extruded
2-9
into the shape of the character. The shaped beam is then returned to
the axis of the tube by deflection electrodes and deflected to the
desired position on the face of the tube.
Random display rates of 50,000 characters per second are possible
with this technique. The limiting factor is the time required to
position the character on the face of the tube rather than the time
required to shape the beam. The Charactron tube is not limited to
the generation of alphanumeric characters but can also generate any
symbol fabricated in the mask. A typical mask has 64 different characters
or symbols, but ]kk symbol masks have been used, and several hundred
are considered possible.
The Charactron is claimed to have three major advantages over the
stroke or dot matrix method of symbol generation:
1) Rel iabi 1 ity
2) Legibility or definition
3) System complexity
Since the charactron is basically a cathode ray tube, it can be
operated as a conventional cathode ray tube to generate graphical data
and target traces in real time. Any shape and size of symbol can be
chosen since this is a function of the fabrication of the mask. Charactron
tubes are useful for image generation in photographic projection systems
for large-screen displays as well as for direct viewing in console
displays. Recent development permits the simultaneous generation of
alphanumeric and real-time video information by the use of two electron
guns in the tube. Another recent development provides a rear window in
the tube so that a photographic image can be projected through the window
and superimposed on the face of the tube with the electronically generated
picture. Fiber-optic face plates have been used to avoid parallax by
bringing the image from the inner surface to the outer surface of the
tube.
2-10
2.2.5«3 Film or Photographic Projection Systems
Large-screen display systems based on projection of photographic
images have been used in a number of existing Command and Control Systems
6 7 8
and several specific systems have been described in the literature ' ' .
In essence, these systems involve:
1) A symbol generator for converting the digital information
to a shaped symbol or character on the face of the CRT
2) An image generator for positioning the symbols and generating
graphical data on the face of the CRT
3) Processing equipment for exposing film to the image on the
CRT, for developing the film, and, if necessary, for making
pr i nts
k) Slide or film storage and selection equipment for storing
the film images and making them available upon call
5) A projector and screen for projecting and displaying the
selected image.
Usually, a multiple projection system is used to permit the simul-
taneous projection and superimposing of multiple images to generate
multi-color displays or to superimpose multiple overlays over a map
background. Systems that superimpose three or four independently
selected images encounter difficult registration problems in the
final projected display. Other systems that contain the multiple images
on a single film chip overcome the registration problem, but the image
size is reduced and flexibility in selecting the combination of images
to be displayed simultaneously is lost. A more recent development
proposes the use of. a lenticular type film in which three separate
color images are contained on the same film image.
2-1 1
A number of the photographic projection systems in current use
employ discrete slides as described above, but a few use a continuous
film strip to provide more rapid updating of the display and to permit
a simpler mechanical system than one in which individual slides are
selected independently. The flexibility offered by random slide
selection is sacrificed. The continuous film strip type projection
system is more suitable to rapid updating of pseudo-real-time displays
where the same type of information is displayed continuously but
updated rapidly. The individual slide approach is more suitable to
situation displays where a large number of different kinds of situations
or pictorial combinations are available, any of which may be required
at a given time and in any sequence.
The photographic projection type systems are currently the most
practical solution to a large-screen display where continuous operation
is required. However, because of the relatively slow response time,
the inability to display dynamic information, and the mechanical
equipment involved, this is not a desirable long range solution. It
is believed that this type of system will be obsolete before the 1970
period and should not be considered for a 1970 system.
2.2.5.^ Photochromic Display Systems
The use of photochromic materials offers considerable promise for
9
future display systems . Photochromic materials are organic dyes which
become opaque when exposed to ultraviolet light, and return to the
transparent state when exposed to heat or infrared light. By coating
a transparent film with a thin layer of photochromic material, a
"photographic" type media can be produced in which the chemical process
is reversible. An image can be exposed with ultraviolet light and
erased with infrared light.
2-12
The exposed image will decay at room temperature at rates depending
upon the particular chemical compound. Typical persistency times for
photochromic materials used in display systems range from approximately
2 seconds to 15 minutes. Faster decay times can be obtained but this
is usually not desirable for display purposes. Achieving longer
persistence times would probably require cooling the image since the
photochromic decay is inhibited by cold temperatures.
Photochromic materials exhibit a fatigue characteristic at present,
after a few hundred cycles of a particular spot. Red, blue, or green
colors can be obtained with a resolving power capability in excess of
1,000 lines per millimeter. The sensitivity varies with the photo-
chromic material but is about 1/3-watt-second per square centimeter.
The persistency of the image can be controlled by varying the tempera-
ture, the material, or the method of applying the material to the
base. The earlier photochromic display systems generated a dynamic
display by focusing an ultraviolet light through a lens system onto a
photochromic film; the u 1 trav iol et light being mechanically positioned
by a servo-mechanism. Since the photochromic material becomes opaque
at the point at which the ultraviolet strikes, projection type displays
can be generated by inserting the photochromic material between the
lamp and the lens of a projection system. Moving the lens through
which the ultraviolet light is focused causes the opaque spot on the
photochromic film to move, generat i ng a dynamic display on the screen.
Shining an ultraviolet light beam through a character-matrix mask
permits the generation of alphanumeric characters on the display
screen. Special symbols can be generated in a similar manner. This
type of display is interesting for tracking a limited number of targets
or for generating displays that change relatively slowly. However,
the speed of the photochromic material and the mechanical motions
involved in deflecting the ultraviolet light limit the speed of this
type of device.
2-13
In a newer development, a cathode ray tube is combined with the
photochromic film to permit the electronic generation of an image. In
this development, a fiber-optic face plate cathode ray tube is used
to generate an image on the outer surface of the face of the cathode
ray tube by conventional techniques. The ultraviolet light from the
phosphor on the inner surface of the face of the cathode ray tube is
transmitted through the fiber-optic face plate to generate an opaque
image on the photochromic film. A dichroic mirror that transmits ultra-
violet light and reflects visual light is sandwiched between the fiber-
optic face plate and a photochromic film. Visual light from an external
source is projected through the photochromic film onto the dichroic mirror
which reflects it back to a viewing screen. The opaque image on the
photochromic film prevents the light from the projector from striking
the dichroic mirror. Hence, this Image is reflected onto the screen.
At the present time, the speed of photochromic materials limits
the character generation rate to approximately 10 characters per
second in this type of display. If work on faster photochromic
materials is successful, this approach could provide an attractive
al 1 -el ectro-optlcal dynamic large-screen display with no mechanically
moving parts. Photochromic display systems combining electronic,
photochromic, and projection techniques are very promising for use
in a 1970 system.
2.2.5.5 Light-Valve Systems
The term light-valve In a generic sense refers to any system in
which light passing through the system Is modulated. However, the
term Is usually used In a narrower sense to refer to a cathode ray
tube projection display system using a Schlleren optical system.
In a typical system of this type, a metallic mirror-like surface
covered with a thin film of oil Is placed Inside an evacuated cathode
ray tube type device. An electron beam Is used to generate an Image
2-]k
on the oil film. This is similar to the operation of a normal cathode
ray tube except that the image is generated on the oil film rather than
on a phosphor face. The electrons impinging on the oil film create
electro-static forces that cause a temporary deformation of the oil film.
When a high intensity light source is focused on the oil film, the
light is reflected at a different angle for those areas that have
been deformed by the electron beam than for the remainder of the oil
film. Passing the reflected image through a ladder-like grating permits
selective passing of the light, depending upon whether it was reflected
from a deformed area or a non-deformed area of the oil film. Hence,
the desired image Is displayed on the viewing screen.
Light-valves are promising for future display systems and will
probably be in widespread use in 1970. At present, they suffer from the
severe disadvantage of short cathode life (20 to 200 hours MTBF) . Since
it is necessary to have an oil film inside the vacuum, it is difficult
to maintain a good vacuum. As a result, there is a tendency for the
cathode to be poisoned by evaporated oil. Light-valve systems of this
type are In common use in large-screen theatre-television systems.
However, these systems are operated for short periods of time for
special events, and considerable time can be allotted prior to the
event for bringing the system up to proper operation. Unfortunately,
in the military command and control environment, the system is required
to be in almost continuous operation. Another disadvantage is that
multi-color displays require multiple projection units with a consequent
registration problem.
Considerable development efforts are being expended toward improving
the performance, reliability, and life of light-valve systems. The
Rome Air Development Center, in particular, is sponsoring extensive
efforts toward improving light-valve systems. It is their belief
that light-valve projection systems will constitute the next generation
of large-screen display systems^ It Is likely that projection light-
valve systems will constitute the next generation of large-screen display
systems, but this Is an interim solution. Such systems will probably
be surpassed by other techniques for a system designed for 1970.
2-15
2. 2. 5^6 Electroluminescent Displays
The major appi i cat ions of electroluminescent materials in display
equipment so far have been in the form of individual character or
symbol indicators . In these devices, each character position in an
alphanumeric display is represented by an electroluminescent panel
which can be caused to display any one of a predetermined set of
characters depending upon the electrical signals applied to the
device. However, extensive research and development efforts have
been devoted to the use of electroluminescent materials to fabricate
a complete display screen capable of displaying graphical data as
well as alphanumeric characters.
Electroluminescent displays offer the advantages of an all-solid-
state display without moving parts or projection optics, a flat display
requiring very little depth, and sufficient brightness for viewing under
normal ambient room lighting conditions. An electroluminescent element
consists of a thin layer of phosphor powder that is embedded in a
dielectric medium and sandwiched between two parallel plate electrodes,
one of which is transparent. The application of an alternating voltage
to the electrodes causes the phosphor to emit light.
Aside from the discrete character display, the electroluminescent
display which has been developed further than others to date has been
12
the electroluminescent crossed grid display . This display uses a
continuous electroluminescent sheet with the electrodes on one surface
subdivided into parallel strips in the X direction and with the
electrodes on the other surface subdivided into parallel strips in the
Y direction. Applying excitation to an X and a Y strip will cause
the electroluminescent material to emit light at the intersection.
In this XY selection, each wire carries approximately one-half the
2-16
voltage necessary to excite the electroluminescent material so that full
excitation voltage occurs only at the intersection. A continuous sheet
of non-linear resistor material is coated on the electroluminescent
material between two sets of electrodes to avoid partial excitation
and partial light at points along the selected X and Y strips other
than the point of intersection.
This approach is useful for a large-screen or console type display.
Real-time dynamic displays, such as target tracks, can be generated by
properly sequencing the selection of X and Y grids. Alphanumeric
characters and symbols can be drawn on the same display. However, it
is necessary to regenerate each spot on the display periodically since
it has no storage characteristic. As a result, this type of display
requires either an external storage or computer controlled regeneration.
To avoid noticeable flicker, the picture must be regenerated at least
30 times per second. The frame rate of 30 per second, and the fact that
1 - 5 microseconds are required to energize each spot on the display,
limit the total number of positions that can be activated. Periodic
action is required for active spots that remain static as well as for
those that are changing.
One display of this type provides a 256 x 256 matrix in a 16 x 16
inch display panel. This display panel is 32 inches thick. The spot
size is approximately 1/10 of an inch. It is expected that spot sizes
of ]/kO to 1/50 of an inch are realizable in the near future, and that
1/100 of an inch is feasible.
In another type of electroluminescent display, a continuous sheet
of electroluminescent material is deposited over a sheet of p iezo-el ectr i c
13
ceramic . With the proper voltage applied to the electroluminescent
material, a mechanical shock wave travelling through the p i ezo-electr i c
ceramic can generate sufficient voltage to energize the electro-
luminescent material in the vicinity of the shock wave. Introducing
2-17
a shock wave to one edge of the ceramic causes a light signal to propa-
gate across the electroluminescent material as the shock wave propagates
across the ceramic beneath it. A reduced shock wave on one edge, combined
with a shock wave on a perpendicular edge, can cause a point of light
corresponding to the intersection of the two wave motions to propagate
across the display. A non-linear resistor material is again used to
suppress partial excitation. Controlling the time of the two shock
waves provides the ability to position the spot of light as it moves.
A third approach to electroluminescent displays can provide a
dynamic large-screen display that does not require periodic regeneration.
In this approach, the display screen is fabricated with a matrix of
discrete electroluminescent elements, each having an associated storage
element. An XY selection matrix is used to energize a specific electro-
luminescent element. The associated storage element then maintains
the electroluminescent element in that state until it is cut off by
another XY selection operation. At present, the addressing rate is
limited to a switch-on time of approximately 10 microseconds per element.
The switch-off time is approximately 30 microseconds, but it is not
necessary to maintain the electrical signal for this length of time,
it is anticipated that the switch-on time can be reduced to 5 micro-
seconds in the near future. Resolving powers of 10 lines per inch can
be realized nowwithlS - 20 lines per inch considered feasible in 1970.
This approach provides a true dynamic large-screen display with
exact registration and positioning without mechanically moving parts
and without an optical projection system. Since the individual
storage elements eliminate the necessity for periodically regenerating
the picture, only those elements that change must be activated and
energized or de-energized.
A multi-color display would be difficult but is conceivable by
segmenting each element of the display into three elements corresponding
to a three color system. This type of display would be quite expensive
due to the electronic selection of individual elements and the electronic
2-18
storage associated with each element. However, it is a practical display
in that a dynamic large-screen display of this type can be built in a
relatively short time with a high assurance of success. Future develop-
ments in integrated circuit techniques may lower the cost of the electronic
elements sufficiently to make this approach attractive for a 1970 system.
2.2.5.7 Opto-Magnet i c Displays
A different approach to solid-state displays is based on the
magnetic properties of certain thin-film materials that affect their
reflection of light. If a thin-film of magnetic material of this type
is deposited on a substrate, areas that have been magnetized will reflect
light in a different way than other areas of the film. An XY matrix
selection can be used to generate a magnetic image on the surface. If
a high intensity light is projected on the magnetic film, a visual image
will appear as the result of the effect of the magnetic image on the
reflection of the light.
Contrast ratios of 75 to 1 have been obtal ned, prov id i ng a good display
under normal ambient light conditions. Only a few percent of the in-
cident light is reflected. Resolutions in the order of 5 mils have
been obtained in the laboratory. The intensity varies with the viewing
angle but there is very little variation within angles of approximately
9o^
This is an interesting approach to a dynamic large-screen display,
but it is too early in the development stage to determine with confidence
whether it will be available and feasible for a 1970 system.
2.2.5.8 Crystal Light Modulators
A projection display device using a birefringent KDP crystal
14
element has been proposed . An electron beam in a cathode ray tube is
used to control the passage of light through a KDP crystal in the face
of the tube. This permits a system in which a polarized light is
projected through a rear window in a cathode ray tube and then through
2-19
the crystal modulating element in the face of the tube and onto a screen.
The electron gun in the cathode ray tube generates an image on the crysta
modulator, the polarized light passing through the modulator then
projects this image onto the screen. This approach is being followed
with interest, but there is no indication at this time as to whether
it will be feasible for a 1970 system.
2.2.5.9 Laser Luminescent Display Systems
Conceptually, a large-screen display can be generated by writing
on a luminescent screen with a laser beam. This would be somewhat
equivalent to an "outdoor" cathode ray tube in which the laser beam
replaces the electron beam and the luminescent screen replaces the
phosphor coating and the face plate of the tube. This approach would
offer an advantage over a cathode ray tube in that a vacuum is not
required and a large-screen image can be generated directly. The
difficulty lies in the deflection of the laser beam. However, a number
of development efforts have been aimed at this problem with some
laboratory success . An operational system of this type may not be
developed by 1970, but it offers long range promise. Continuing efforts
in the development of improved lasers and advances in the ability to
deflect laser beams will contribute directly to the ultimate success of
this type of display.
2.2.6 Display Availability in the 1970-80 Period
The investigation of display systems has not progressed far enough
at this point of the study to permit a complete determination of the
availability of different types of display systems in 1970. From the
investigations to date, It is believed that film-based photographic
projection systems and mechanical inscribers will be obsolete. Light-
alve projection systems will be In widespread use but may be phasing
out by 1970. Cathode ray tubes will continue to be a dominant factor
In the generation of displays and In console viewing screens.
V
2-20
Electroluminescent and photochromi c systems offer promise for 1970-80
systems. Ferroelectric and opto-magnet i c displays offer possibilities
depending upon the success of current development efforts. Laser
display systems appear to offer great promise but significant research
and development efforts are required.
2,2,7 Limitations of Present and Planned Displays
Most of the present large-screen displays are limited by the use
of electro-mechanical film based projection systems. Photochromic and
electroluminescent displays are currently limited by the rate at which
individual positions can be altered. Light-valve systems are limited
by short cathode life. It is not possible at this point of the in-
vestigation to discuss the limitations of future display systems.
2c2.8 Recommended Developments to Meet ANTACCS Needs
Development efforts sponsored by the Navy to meet ANTACCS needs
for the 1970-80 period should be concentrated on solid-state techniques
that are adaptable to both fabrication methods. Electroluminescent,
opto-magnet 1 c and laser displays appear to be the most fruitful areas
for development efforts pointed toward the 1970-80 period.
2.2.9 The Evaluation Criteria Recommended
Characteristics and parameters to be considered in evaluating
display systems should include the following:
Display technique
Display media
Console or large-screen
Stat i c or dynami c
Hard copy
Leg i b i 1 1 ty
Color
Br i ghtness
Screen Size
Resol ut ion
Frame generation rate
Response time
Character generation rate
Character generation technique
Image storage capability
Image storage method
Storage and regeneration requirement
Capac I ty
2-21
Character repertoire
Character s i ze
Symbol shapes
Accuracy of position
Reg i s t rat i on
Stabi 1 i ty
Col or capab i 1 I ty
Cont rast
I mage qua I i ty
Opt i ca 1 qual i ty
Processfng requirements (if any)
Image handling requirements
Accessab i 1 i ty
Background illumination permissible
Vi ewi ng d i stance
Physical space requirements
Wei ght
Power Requirements
It will not be necessary to make detailed comparisons and evaluations
of each of these characteristics since many of them are common to most
display types. Such characteristics would be used to rule out a limited
number of displays that do not posses the characteristic. Other charac-
teristics might be common for most displays but a unique property of a
specific display technique could enhance this characteristic to offer a
strong advantage to that technique. Those characteristics that vary
significantly from display to display, but within acceptable limits, will
be compared to permit evaluation of the characteristics of acceptable
display technologies.
2.2.10 Conclusions and Recommendations
Only brief preliminary conclusions and recommendations can be made
at this time. Emphasis should be placed on the development and use of
solid-state displays that do not require periodic regeneration and on
batch-fabrication of display screens and arrays.
2-22
References: Displays, Section 2.2
1 "Physical Principles of Displays - Classification,"
H. G. Talmadge, Jr., Electronic Information Display Systems ,
Spartan Books, Washington, DC, 1963, pp. 69-86
2 "Cathode-Ray Tubes," F. R. Darne, Electronic Information Display
Systems , Spartan Books, Washington, DC, 1963, pp 87-109
3 "Criteria for Group Display Chains for the 1962-65 Time Period,"
Technical Documentary Report No. RADC-TDR-62-3 1 5 , Rome Air
Development Center, July 1962, pp 1-2
4 "A Synopsis of the State of the Art of Dynamic Plotting Projection
Displays," R. Anderson, Second National Symposium of the Society
for Information Display, New York, October 1963
5 "Advanced Display Techniques Through the Charactron Shaped Beam
Tube," J. H. Redman, Society for Information Display Symposium,
March 1963
6 "Colordata Display" Hughes Aircraft Co. Brochure, Fullerton,
Cal i fornia, 1 963
7 "Artoc Displays," R. T. Loewe, Electronic Information Display
Systems , Spartan Books, Washington, DC, 1963, pp 231-246
8 "DODDAC - An Integrated System for Data Processing Interrogation and
Display," W. F. Bauer and W. L. Frank, Proceedings EJCC, Washington
DC, December 1961
9 "Photochromi c Dynamic Display," E. J. Haley, Electronic Information
Display System , Spartan Books, Washington, D.C., 1963, pp 110-120
10 "Epic Display", H. L. Bjelland, Proceedings 3rd National Symposium
on Information Display, San Diego, Calif., February 1964, pp 286 - 299
11 "Display Applications of Electroluminescence," M. S. Wasserman,
Electronic Information Display Systems, Spartan Books, Washington
DC, 1963, pp 121-125
12 "Non Linear Resistors Enhance Display Panel Contract," H. G. Blank,
J. A. O'Connell, and M. S. Wasserman, El ectron i cs , August 3, 1963
13 "Solid State Display Device," Stephen Yardo, Proceedings of the IRE,
December 1962
14 "Solid Crystal Modulates Light Beams," E. Lindberg, Electronics ,
Vol 36, No. 51, December 20, 1963, pp 58-51
15 "A Fast, Digital -Indexed Light Deflector," W. Kulcke, T. Harris,
K. Kosanke, and E. Max, I BM Journal , of R & D , Vol 8, No. 1,
Jan 1964, pp 64-67
2-23
2.3 DISPLAYS--USER TECHNOLOGY AND SOFTWARE
2.3.1 I nt roduct i on
2.3.1.1 Objective
Real time data processing systems have become an important point of interest
The development of digital transmission systems and the availability of bulk
data storage devices have stimulated the use of on-line systems in which
information is entered into the data processor as it is generated , and outputs
are requested from the computer as they are required , and in fact limited to
that information needed at the moment. The on-line concept established a
requirement for an intimate relationship between man and compu ter--one in which
not only the characteristics of the computer are important, but where equal
concern must be given the communication devices by which man interacts with the
system. This study is concerned withthe application of on-line displays to
military command and control.
2.3.1.2 Historical Review and Perspective
Perhaps one of the first display devices associated with a computer system
was the simple cathode ray tube (CRT) display. For example, such a device was
available In 1953 on the ILLIAC (University of Illinois) computer where two
tubes were driven in parallel. One CRT was mounted for visual observation
whereas the second was associated with a camera capable of photographing the
computer generated display. The computer controlled the film advance. While
the primary use of this device was the rapid generation of graphic information,
another use was the on-line monitoring of the progress of a calculation by the
programmer. By appropriate displays, he was able to detect programming errors
or, during production runs, make better initial guesses during iterative
procedures or parameter studies. This was perhaps one of the first on-line
display devices.
2-24
It is important to define what constitutes an on-line display device.
From many points of view, all of the following can be considered as display
dev i ces :
Typewr i ter
Plotter
Pri nter
C 1 osed c i rcui t TV
Document viewers
CRT consoles
For purposes of this investigation, however, we limit considerations to
essentially two types of devices, the single operator console and the group
display. These must satisfy the following criteria:
1) Directly tieable to data processing system.
2) Ability to initiate messages or control signals from a data
entry keyboard or switches for transmission to the computer.
3) Ability to receive messages or control signals from the computer
and display them to the operator or viewer.
"Off-line" devices are included if they receive information which is computer
generated. Conventional printing and plotting equipment, however, are excluded.
So are document viewing devices associated with the information retrieval
problem. These devices generally operate on reference libraries and not on
digit ized data.
Based on the above, the typewriter station, the keyboard with CRT console
and the large screen viewing system are the basic items applicable to this
study.
2.3.1.3 Typical Systems and Operating Modes
SAGE
Perhaps the first, or at least most well-known display system was that
associated with the SAGE System. The operating principle in SAGE Direction
Centers is the interplay of man and computers via display consoles for the purposes
of making a composite number of simple decisions concerning the air threat at
any instant in time. To do this, geographic information is presented by the
computer to the console operator on a volatile CRT screen, and the human
responds to the machine by operating a light gun and button keyboard. Alarms
2-25
and alerts, In both audible and visual form, are available. In a typical
center, there are close to one hundred consoles. Depending upon the function,
the console features differ from one another, there being over a dozen "special
purpose" configurations. Operating characteristics have been cited where the
computer reads up to 5000 console switch actions every 2.5 seconds. During
this period, 200 different displays may be generated consisting of 20,000
characters, 18,000 points and 5000 lines.
Typically, a console has a 19 inch Charactron tube and a 9 inch Digitron
tube manufactured by General Dynamics/Electronics and Hughes respectively.
The larger scope is used as a situation display capable of showing alpha-
numeric characters and lines, whereas the small CRT is a data summary display
capable of only alphanumeric information. Whereas the former leads to con-
siderable flicker to the casual observer, it is maintained that operational
personnel who are subject to a special environment of blue light find no
problem in working with the displayed Information.
In the SAGE System, on-line devices first came to be used on a large
scale. The significant application principle here Is the use of the console
in the area of computer assistance where human judgment can be applied.
NORAD
Another area of Interest in the application of display techniques and
devices is NORAD.
The current NORAD Complex at Colorado Springs (apart from the plans of
425L) Include two major installations, the Space Detection and Tracking System
(SPADATS) and the Combat Operation Center (COC) . The display devices and
techniques used here are:
2-26
SPADATS - a) 5 ft. x 6 ft. wall map of the world with pins shovnng
locations of sensors, communication elements, and data
processing stations.
b) Tote board presenting tabular information of all space
vehicles currently in orbit. Manual updating of perigee,
apogee, period, etc. is performed daily.
c) C losed c i rcu it Wollensak TV for' transmitting parts of the
display on the tote board to other operation rooms.
COC - a) Large screen of the North American continent capable of
showing tracks. This system uses the Iconorama projection
system which automatically updates film chips from tele-
type messages.
b) Smaller screen devoted to showing the BMEW s system on a map
background of the Arctic region,
c) Weapon Status Board - registers which display the status
of forces.
Of considerable interest here is the Iconorama system which is perhaps the
first on-line, multi-color, group display system to be installed. The
NORAD System generates information which can be displayed in the following
add i t i ona 1 s i tes :
Joint War Room, Pentagon
Air Force Command Post, Pentagon
SAC Command Post, Offutt Field
Canadian Joint Chiefs of Staff
National Resource Evaluation Center
Air Force Command Post
The display activity to data at the Air Force Command Post has been
limited to several rear projection screens capable of showing slides and
films. Perhaps the most dynamic display is the Iconorama System which is
fed by NORAD. In addition, there are status boards exhibiting the defense
conditions (DEFCON). Also there is a Bomb Alarm display which consists of
a map of the United States consisting of colored regional blocks. Upon an
indicator going off, an appropriate light goes on behind the map and lights
up the endangered area.
2-27
During the spring of 1962, this system was augmented by the first stage
of the 473L program called the Operational and Training Capability (OTC)
phase. This was implemented by IBM Federal Systems Division by the Intro-
duction of the IBM 1401/1405 (disc) systems together with the DC400B/DIB
display and interrogation system of Thompson Ramo Wooldridge.
The latter system consisted of two RW consoles having single 10-Inch
CRT displays together with a sophisticated keyboard. Of these, one console
was a remnant of earlier equipment while the second console was a newly
manufactured copy. Both consoles were interfaced with the 1401 by use of a
Display Interface Buffer which is a core storage device. These consoles were
to be used In the Command post by placing them at the disposal of the various
area desks as an on-line tool for Information retrieval and analysis. The
following functions were selected for this initial application:
Emergency Actions
Defcon Actions
Plan Abstracts
Fl I ght Fol 1 owl ng
Status of Forces
Ai rf ield Fac i 1 i t les
Aircraft and Missile Characteristics
The purpose of the OTC was twofold: to achieve some automated capability
rapidly; and to experiment with equipment and techniques In anticipation of
the next phase of 473L development. The latter point motivated the interest
in on-line display devices so that experience would be obtained for design of
the Interim Operational Capability.
4) Defense Communication Agency : Defense National Communication Control Center
The Defense National Communications Control Center (DNCCC) Is the focal
point for the controlling and supervising function of the Defense Communication
Agency over the Defense Communication System. The latter encompasses all of the
telecommunication requirements for the Department of Defense. The basic function
of the DNCCC Is the maintenance of world-wide communication traffic status. In
this capacity three basic on-line displays are generated to show operational
conditions; traffic status, system status and read-out panel.
2-28
These displays are computer driven, wall-lined panels. The first tv/o
ewe static in composition, that is, they are fixed format displays on v;hich status
is demonstrable by the manipulation of colored lights as generated by the Philco
2000 computer system.
The System Status panel is an 8 ft. x 15 ft. map of the viorld on v/hich are
etched major trunk lines and system relay stations for which the back lighting
can be red, green or yellow.
Another panel is the Traffic Status Board which is an 8 ft. square display.
It consists of four bays of nine columns each, showing station backlog status
by use of coded illuminators opposite each identified station.
Finally there is a readout panel which is a 7 ft. square rear projection
display. This display can present tabular and textual information generated
by the computer as a result of an inquiry. The process incorporates a modified
Anelex High Speed Printer which generates text on special material from which
a film transparency is generated by non-chemical means. This film is
sequentially added to and is, therefore, spool fed and wound with 70 characters
per line at 100 lines per minute. The letter image is 1 inch high, and is large
enough to be read at 25 ft. distance in a room with ambient lighting of 25 ft.
lamberts .
These three display devices are tied together through a System Supervisor
Console which is under the control of the human operator. The console serves both
to initiate queries of the system and to display results via the station
lights, typewriter, and printer stations.
In addition to the above, there are three special Network Boards Including
the station and trunk conditions for the CRITICOMM Network, the JCS Voice and
Teletypewriter Network and the JCC Network,
Operational experience has shown that the read-out display has found very
little use. This is probably due to the lack of requirement for large group
display of textual material.
2-29
The DNCCC represents stage one of the development of the Defense
Communication Control Complex.
The current DCA procurement is expanding the Control Central Complex by
adding to the DNCCC four area control centers, called Defense Area Communication
Control Center (DACCC) , and five regional control centers, known as Defense
Regional Communication Control Centers (DRCCC) . Those centers will perform
on a decentralized basis, many of the same kinds of functions performed at the
higher level DNCCC. In particular, they will be more responsive to local
problems and system management.
The display requirements for the DACCC have been spelled out considerably
and serve as an example of what is currently demanded. Specifically, a visual
display is specified which will exhibit "in a completely unambiguous manner"
current operations! status. To this end the following guide lines are parameters
by which a responsive display system was to be proposed:
a) Alteration: Ability to make changes to the display within
4 hours per unit change and without removing the
display from normal service,
b) Expansion: Have a residual capacity to allow a 25% increase in
information displayed.
c) Leg i bi 1 i ty and
Resolution: Permit personnel with 20/20 vision and normal color
perception to comprehend the large wall displays and small
individual displays at 15 ft. and 30 Inches respectively.
d) Information Of twenty information items, eight are semi fixed and
Content: 12 are variable. The average frequency of change for
these items varied from a low of 15 minutes to hourly
for 13 items, and infrequently for the remainder.
2-30
2.3.2 Application Technology in Command and Control
The technology of integrating man/machine communication devices must
take into consideration both programming and user techniques. The former
is concerned with achieving system operation and the latter with the design
of proper operating procedures.
Always the on-line application and problem characteristics must be
understood to arrive at an economical system,
2.3.2.1 Man/Machine Coordination
There is close interrelationship between man and machine in a command
and control environment. During both pre-attack and post-attack times,
the machine receives information from communication devices and displays it.
The man analyzes the incoming information as to the status of his forces and
those of the enemy. The machine receives additional information and updates the
situation display. The man evaluates the military situation based on the
information the machine has displayed. In so doing, he makes requests of the
machine to which the machine responds, since it is impossible for the system to
cover, under normal output procedure, data reflecting all contingencies in which
the commander might be interested. The man identifies certain courses of
action, and the machine computes hypothetical effects based on the various
possible choices. The man makes the decision and the machine communicates
the commands and records them.
The function "computes hypothetical effects" is probably futuristic.
At the present time, there is little real-time or operational war gaming
capability in command systems to help the commander make his decision during
the actual post-attack period.
Figure 2-1 refers to the overall command function. There are obviously
many support functions which require close man/machine coordination.
To fix the user concepts, it is convenient to illustrate by an actual
example indicative of a commander's requirements. Consider the problem of
interrogating the information files in a system. If these files contain status
information regarding forces and resources, then some typical interrogations
of the system might be:
2-3
MAN
/
MACHINli
,/■'
/
Receives and Displays Information
Aiialyz
es Status of Forces
/ ■ ' /
Updates Situation Display
Evaluai
es Military Situation
Responds to Data Requests
/
Identif
les Courses of Action
'
Computes Hypothetical Effects
Makes
Decision
/ / /
/
/
Communicates and Records C
ommands
Figure 2-1 MAN/MACHINE COORDINATION
2-32
1) Tabulate all POL facilities on Russian and Chinese Bomber Air
Bases whose capacity is greater than 500 metric tons of jet fuel.
2) List all NATO air defense bases with a probability of survival
grea tec than . 85 .
3) List all Navy Bases with a greater than 757, probability of survival
and whose residual capacity of P02 is greater than 20X.
The conventional procedure in fulfilling these requests is for the consumer
to fill out a request form and then await the manual and machine steps, shown in
Figure 2-2;,to be completed. The computer Input format required in step 4 typically
demands a trained specialist to convert the free text and terminology used by
the requestor to machine understandable nomenclature and fixed form. This
process requires table look-ups and transcription from code books, indexes
and tables of acceptable terms.
Disadvantages encountered In this process are:
1) Need to carefully adhere to spelling and to form; for example,
abbreviations, plurals, possesslves, etc. may be excluded.
2) Use of special words. Synonyms may be prohibited. The
terminology of users varies so that a common vocabulary
acceptable to all is impossible.
3) Punctuation. The compounding and marking off of segments
of the query may lead to logical errors.
4) Need to learn special rules and codes. Change of codes will
affect all users at the problem originating level.
The on-line display console permits a short circuit of almost all of
these steps by providing the consumer with a direct entry on the files. Not
only are time delays eliminated by the simple brute force approach of by-passing
at least eight manned stations, but some of the Inconveniences of maintaining
security and generating possible errors at each point are also avoided.
2-33
Fill out request
form using
appi i cat ion-
oriented language
Submi t to
processing center
I
Receive message
and log
I
Translate to
Computer -or i en ted
Language and Format
i
Key Punch and
Verify
I
Run on
Computer
Output
Control
Log Output
and Dispatch
i
Submit to
Consumer
i
Study
Resul ts
Consumer
Messenger
Dispatcher
I nformat ion
Special ists
Key Punch
Operations
I nformat ion
Special ist
DI spatcher
Messenger
Consumer
Figure 2-2
OVER-ALL COMMAND FUNCTION
CONVENTIONAL PROCEDURE IN REQUEST FULFILLMENT
2-34
The advantages of this approach are:
1) Direct consumer/system interface
Fast service
No intermediaries that can cause errors or delays
Complete control
Remote opera t i on
Di rect response
2) User oriented language
No code books
No d i ct i onar i es
No syntax rules
3) Error checking and control
The procedures that may be employed with the displays under discussion
here are illustrated in the following example. Assume that the display
console keyboard includes the labelled buttons:
Military installations
Geographi ca 1 1 imi ts
Pol i ti ca 1 limits
Owner limits
User limits
Current totals
Amount degraded
Residua'l capacity
Hard- copy output
CRT output
The steps in entering query (1) are:
1) Pressing military installations, the display given in the
Figure 2-3 is presented. The operator selects AIR BASES .
2) The selection of one item, AIR BASES, causes a second display
to appear automatically as shown in Figure 2-4. The operator
selects BOMBER .
3) The selection of BOMBER causes a third display to appear as shown
in Figure 2-5. It represents the possible choices in selecting
attributes about Bomber Air Bases. The sample query dictates that
the operator choose FUEL STORAGE.
2-35
MILITARY INSTALLATIONS
ALL
MISSILE BASES
AIR BASES
NAVAL PORTS
AIR DEFENSE
C AND C CENTERS
DEPOTS
ARSENALS
AIR BASES
_ ALL
_ TANKER
_ BOMBER
_ FIGHTER/BOMBER
_ FIGHTER
_ ALTERNATE BOMBER
_ ALTERNATE FIGHTER
_ EMERGENCY RECOVERY
EMERGENCY FIELDS
FIGURE 2-3
LIST DISPLAY OF MILITARY
INSTALLATIONS
FIGURE 2-4
LIST DISPLAY MODIFYING
MILITARY INSTALLATIONS
FACILITIES
MISSILE LAUNCHERS
RUNWAYS
MISSILE STORAGE
NUCLEAR WEAPONS
MISSILE CONTROL C.
FUEL STORAGE
KEY COMMUNICATION
KEY TRANSPORTATION!
FIGURE 2-5
LIST DISPLAY MODIFYING BOMBER
AIR BASES
FUEL STORAGE
ALL
AVGAS I
AVGAS II
JET I
JET II
^DIESEL
MISSILE LIQUID
MISSILE SOLID
FIGURE 2-6
LIST DISPLAY MODIFYING
FUEL STORAGE
2-36
4) The next d i sp lay , shown in Figure 2'. 6; mod i f i es Fuel Storage and two
selections are made, JET I and JET I I , completing this sequence of
d i sp lays .
5) The next function key pressed is one labelled Pol i ti ca 1 Limits .
This causes the display shown in Figure 2-7 to appear. From it,
two selections are made; RUSS I A and CH I NA .
6) Next, Current Totals is pressed, causing a format display to appear
as given in Figure 2-8. The operator will insert 500 in the third
b lank.
Having completed the entry of the query, the operator now selects the
output media. This choice will be dictated by urgency for results, amount of
detail desired, size of list likely to be generated, and desire for permanence
of copy. The selection of either of the output media will terminate the request
procedure and cause the request i tan to be added to the internal processing
queue.
FOLITIGi^L LIKITS
RED
BLUE
NEUTRAL
U.S.
RUSSIA
CHINA
TuEATY GnPS
BLOCS
NATIONS
FIGURE 2-7
,IST DISPLAY, POLITICAL
LIMITS
TOTALS
SELECTED BY
GREATER. .
PER
CENT
LESS
SELECTED BY
GREATER. .
AMOUNT
LESS
FIGURE 2-8
FORMAT DISPLAY, CURRENT
TOTALS
2-37
2.3.2.2 Problem Characteristics
Typical problem characteristics associated with on-line display systems
a re as foil ows :
') Rea 1- 1 ime - this refers to both the performance of the total problem
and the responsiveness of the system to individual operator action.
By definition, since operator response and action is involved, real-
time is measured in human terms. Two different operations are
identified. The first involves entry of information, under program
guidance, to make up a complete message or action request. The
second Is the processor response to the message.
An example of the former is the step-by-step composition of a message
by use of a keyboard or function keys as described in the previous
section. Typically, each key generates one or more characters of
information which are collected, operated on and stored by the computer.
Pressing a key may also cause a display to be generated, providing
guidance for the next stop, or requiring the filling in of further
information. This mode of operation must permit the operator to enter
data at his own speed, which effectively means that he must be permitted
to press keys at a 60 word per minute typing rate or within 200 milli-
second intervals.
The second aspect of response is the fulfillment of a particular
request by the processing central. The total action may Involve:
Request validation
Information retrieval
Information transformation
Data formatting
Output generation
Depending upon the specific application, the response time required
of the system may vary from a few seconds in a command and control
system to several minutes for an inventory search. In one particular
strategic command and control system the response times for fulfilling
requests were specified as shown in Table 2-1.
2-38
1 nd i V i dua ]
G r ou p
Form of Output Presentation
Hard Copy Console
Group Display
10-30 minutes 3- 10 sec.
X 3-10 -ec.
X
30 seconds
TABLE 2-1 - Response Time for Fulfilling Requests
2) Random Transactions - Since the input associated with display
devices is generated by people, and the reaction time to outputs is
dependent on people, the system transactions observe no fixed time
pattern or schedule. Hence the processor must be capable of supplying
random servicing of the on-line stations.
3) Large Storage Capacity Requirement - Systems with on-line displays
invariably involve much information in the form of a data base in
data handling applications, or as a program library in scientific
computing. Otherwise, on-line devices can hardly be economically
justified. Since man and his judgment are involved, there is a
requirement that this information be randomly accessible. Hence
bulk storage media and their efficient utilization are important in
system design.
4) Many Stations - typically, the on-line system will have many trans-
action stations. For example, a logistics system may involve
thousands of inquiry sets, while a CIC will include dozens of con-
soles. Multiple users compete for servicing from the central
processor.
5) Independent Functions - As well as many users operating in parallel,
individual functions or tasks vary from one station to the next.
For example, in NTDS , many different operations are defined, any
of which may be initiated at any of these stations.
6) File Access - The multi-functions and multi-users operate on common
files. Hence, file order and file integrity must be maintained.
The latter is a problem since the functions operating in parallel
can be both extracting information or modifying the data base at
the same time.
2-39
2.3.2.3 The Naval Application
In Figure 2-9 there is shown schematically the data processing system
from the point of view of the commander. He receives information from the
machine; frcMn console displays and group displays, and from hardcopy print-
outs. Console displays reflect the working display to help prepare and format
group displciyb and to ledct to particular requests of the commander. Group
displays represent the major standard output to the commander reflecting tactical
situations, i.e. situations which change rapidly. Hardcopy output represents
the back-up data which is frequently used for reference in this perspective.
The data processing system itself is simply a "black box" from the commander's
V i ewpoi nt .
The most important aspect from the standpoint of the commander is that
he must communicate with the machine In his own language-- the military language.
This implies that the hardware and the procedures providing the Input/output
to him must be so designed to allow this. Communication with the machine in
a programmer's language, totally foreign to the military man, is unacceptable.
There must be an intermediate translation from the military language to the
language of the console displays, group displays, or hardcopy print out.
The 'bommander" in the above paragraphs is used in a generic sense. It
is meant to include the commander's staff activities and all support functions
necessary to the mission.
2.3.3 Hardware Aspects from the User's Point of View
2.3.3.1 Equipment Features
2.3.3.1.1 Display Console Features
To illustrate the principles of the techniques and operating concepts
to be presented '\p this section, it is well to define a typical display console
in terms of the capability and tools it affords the user/operator. The following
features are assumed:
DATA PROCESSING
SYSTEM
L^~— w J
7T*
/ \
o
c
INTEGRATION DISPLAY
CONSOLES
C
A A A
c
A
n
o ^; c
It- 1 i IT
GROUP DISPLAY
MILITARY ANALYSIS
& COMMAND
/
L «r i
V
HARD COPY
Figure 2-9 SCHEMATIC OF COMMANDER'S DATA PROCESSING SYSTEM
N>
I
O
2-41
1) Alphanumeric Keyboard - This consists of a set of keys conparable to
a standard typewriter keyboard. In addition to the letters and
figures, punctuation and special symbols will be included. There are
usually sixty-four possible characters available since 5 bits are con-
ventionally used for symbol representation. Since there are 43 keys
on a typewriter, this implies the need for a shift key or augmented
keyboa rd .
2) CRT - A readout unit capable of displaying a set of characters or
symbols with line drawing as a possible option. There is typically
a one-to-one correspondence between the symbols available on the
alphanumeric keyboard and those that can be generated with the CRT.
3) Function Keys - This set is dedicated to application-oriented
procedures. Single keys may represent a call for an action, or groups
of keys may be tied together to form a message calling for an action.
To make the device general purpose or multi-purpose, It is desirable
to have the significance of these keys vary on demand of the operator.
One convenient way to achieve this variability Is by a replacable mask
or overlay as is done in at least several commercial products such as
the Bunker -Ramo BR 85, IBM 4554 and ITT Integrated Console.
4) Status Indicators - System status, both Internal computer and console,
is shown by status indicators. These indicators may be labeled
neons; their on or off condition show status.
5) Ala rm I nd lea tors - Alarms or error Indications are conveyed by a
set of labeled neons. Buttons may be associated with these lights so
that operator recognition and resetting can be accomplished.
5) Control Keys - These keys are dedicated to general tasks and support
functions by which system control, data entry, and status requests are
made.
2-42
7) Light Gun - This is a hand-held photoelectric device by vyhich
the user/operator can index any symbol on the CRT.
We illustrate typical assignments to some of the panel elements:
1 ) Alphanumeric Keyboard
Al phabet
Numbers
Punctuat ion
Special Symbols
"Carriage" Control
Sh i ft Control
2) Control Keys
Hardware Configuration Control
On-line system Function Control
Queue Control
Display Control
3) Status Indicators
Equipment Status
Queue Status
Processing Modes
4) Alarm Indicators
Data Errors
Procedure Errors
Equipment Alarms
2o3'3-2 Equipment Operating Concept
2c3»3.2.1 Console Operating Concept
The basic operation of the display console is in sequencing through
the Function KeySo A desired action is defined by pressing an individual
key, or more typically, a group of keys in some ordered sequence.
The following events represent a typical operating pattern:
1) Press key i
2) Get positive response that key i was pressed
3) Computer presents a display on the CRT
4) Enter data into display
5) Visually validate Inserted data
6) Make corrections to Inserted data if necessary
7) Signal end of entry
2-43
The actual sequence followed by the operator for a particular key
is shown In Figure 2-10.
V/hi le this sequence of operator events is underway the computer
engages in a number of actions paced by the speed of the operator. These
steps ave shown in Figure 2-11.
2.3.3.3 System Configurations
There are several possible methods for tieing the on-line display
device to the computer. The appropriate method for each application must be
determined during the system design when the system equipment is being
specified. The various configurations are illustrated in Figure 2-12. It
should be noted that for a given system, not all of the configurations
illustrated may be possible. The remainder of this section is a brief
description of each configuration.
Conf i gurati on Descr i pt ion
a) and b) The on-line device Is connected to a buffered or
unbuffered computer I/O channel. Additional devices
may be connected to the same channel up to some
maximum number. Then additional devices must be
assigned to another channel.
c) and d) The on-line device is connected to a separate buffer
unit which in turn is connected to the computer via
an unbuffered or buffered I/O channel. Additional
on-line devices may be connected to the buffer unit
up to some maximum number.
e) and f) The on-line device is connected to its own buffer
unit which in turn is connected to the computer via
an unbuffered or buffered 1/0 channel. Each on-line
device requires a separate buffer unit.
The separate buffer units appearing in configurations c) , d) , e) , and
f) perform such functions as automatic CRT display regeneration and character-
by-character message accumulation for subsequent transmission to the computer
or on- 1 i ne dev ice.
Operational examples of these configurations include the following:
DODDAC uses configuration a) with the CDC- 160 computer. System 473L uses
configuration c) with two consoles and the IBM- 1401- computer. Configuration c)
was used at Ramo-Wooldr idge for the RW-400 computer and for the IBM- 7090 computer
2-44
Modify
I
Press Kev
Computer Presents
a Display on the
CRT
Take Next
Directed
Action
Yes
End of
Sequence
Enter Data
into Display
Signal End
of Entry
OK
Error
Make
Correction
Figure 2-10
Typical Operator Steps in Use of Function Keys
2-45
data
i
Trans
form
and
Place
In
1
Displ
ay
^
Scan alert 1 i nes
(if interrupts are
not ava i 1 ab le)
non
alert
Exit to
■^f^ Interrupted .^g^
Program
alert
SF
I dent ify
S i gnal
con t ro 1
End of Message
i
Va 1 idate and
Convert Input
Data
i
Accumul ate
Input in
Message Buffer
Is message
complete ?
yes
Jl
Place new
request in
queue
Select and
In i t i ate
appropr late
program
Seek and
generate
displ ay
-H
no
Figure 2-11
Computer Steps in Conjunction with Function Keys
2-46
-3 '
c k :
'VJ^
c)
:>
Oa
r)
-->i^^-^
C = COMPUTER
D _ £: ' ! 7 fT C p
OM-L!>:£ DEVICE
U:''3UrFERED i/0 CHANNEL
BUFFERED I/O CISANNEL
FIGURE 2-12
COMPUTER/ON-LINE DEVICE CONFIGURATIONS
2-47
2.3.4 Software Considerations
2.3.4.1 Implementing and Using the System
From the point of view of the user/operator the display subsystem
should possess the following:
1) Standard procedures - It is well to establish clear rules with regard
to console functions and operations so that there is a minimum of
confusion by the operator. This refers to the following items:
a) Maintain uniform use of groupings of keys. If, for example,
the console keyboard has several physically distinct arrays
of lights or buttons then it Is wise to have each unit in the
group . cons i s tent with the function of the group. Thus all alarm
indicators should be grouped together.
b) Define simple steps for using console buttons that are consistent
within each physical panel grouping. Thus the alphanumeric keys
should all cause a character to appear on an associated display
for each key depressed. System alarm indicators should behave in
identical ways for all alarms.
c) Require all data input via the a 1 phanumer ica 1 keyboard to f i t a
prescribed format whenever possible. This has the advantage of
allowing the computer to guide the operator and permit computer
transformation of data from external to internal representation.
2) Ease of Use - Program as much as possible and provide as many
flexibilities as possible. Do not sacrifice user simplicity and
flexibility at the expense of more complicated "one time" programming
3) Flexi b i 1 i ty - Usage of display devices will invariably lead to
improvements in procedure and technique. Hence it will be desirable
to be able to make changes often and easily. This implies providing
the ability to alter individual steps and the logic flow.
4) G rowth - Application areas will grow as new uses are found for the
displays, especially if the devices are general purpose. Hence, it
should be possible to add new functions to the programming system.
2-48
5) User Or i enta t i on - The display subsystem should be as much user
oriented and user understandable as possible. Hence, it is desirable
that the design, implementation, and modification of specific functions
be as much "professional" programming independent as possible.
Optimally, the system should be manageable by the user, once the basic
programming has been done.
The following operating principles should be preserved:
1) Lights that present status should uniformly all present either
negative or positive information whenever possible. This rule may
be compromised for a second rule which states that Indicators should
go In the "on" position for the exception, and not the rule. (This
rule Is, however, often violated by the equipment "on" indicator.
This violation is probably justified) .
2) Whenever a button Is pressed by an operator, some positive and
Identifiable response should be given by the computer to the operator.
Thus, if a control button is pressed, an associated neon should go on
or the display should indicate the action. This redundancy should
be caused by a return message from the computer to serve as an intermodule
communication check (and hence not require hardware error check ).
3) The console buttons should have associated markers to indicate a
functioning keyboard. The operator should be made aware of a full
buffer if keys are pressed too rapidly. This can be served by an
indicator or by locking the keyboard.
2.3.4.2 User/Operator Techniques
2.3.4.2.1 CRT Displays and Their Use
CRT displays may be classified Into four basic types:
1) Message displays, which do not require modification in any manner
by the operator.
2) List displays, which require the operator to make selections from
a prepared list of items.
2-49
^) Forma t displays, which require the operator to make entries in well
specified positions,
4) Free displays, which allow the operator to make entries into any
pos i t i on .
Of these four display types, the first is computer generated (e.g. outputs),
the middle two are computer presented or computer generated for operator use,
and the last is operator generated. Typically, only the first three are
employed, the last being useful only for free text entry.
One of the important features that should be available with CRT displays
to aid in data entry is knowledge of current "position" of the "platum". That
is, the position on the CRT where the next character will be placed. This can
be handled either by hardware or by programming. In either case, some symbol
is dedicated to serve as a'lViarker" having the property that it will always move
one character position to the right (or to the first location of the next row
if it now occupies the last position of a row) whenever an alphanumeric key Is
pressed. The position previously occupied by the marker will be taken up by
the data symbol just entered.
The message display is generally information supplied by the computer
regarding operational status, actual output, and such items as file indices
and reference tables. The free display would be information supplied to the
system for future reference purposes. Also, the free display could serve as the
input for generating certain data bases for say, intelligence files.
The 1 i st and format displays are the work horses of the system. The
former consists of a list of items from which the operator may choose in
making parameter selections. By use of special selection keys, the operator
is able to select any number of items from the list. Furthermore, choice of an
item from one list could lead to the presentation of a second list, permitting
further choices. Such an arrangement leads to the concept of multi- level
sequencing through an indentured index taking the analyst from the general
category to the specific by the most direct route. For example, if the
operator had selected Air Bases from the choices open to him, then the
2-50
system may next present him a breakdown of this category as was illustrated.
This process could, of course, continue to any level. The item chosen at the
lowest level reached will then serve as a selection parameter in whatever
action is being generated. In effect, the parameter will be the logical
"and" of this item with all the higher level choices made to get to this
pa rt icu lar list.
In a list display it is also possible to make multiple selections within
a particular level. To establish some bounds to these selections, if a
multiple choice is made, the operator is prohibited from going to a next lower
level. This limitation is imposed to avoid the unlimited handling otherwise
possible. Thus, of the two "trees" shown in Figure 2-14 case (B) is now
a 1 lowed.
The format display facilitates Input of data that is structured. It has
a well established form and the input data must conform to the limits set
by the format. An example of such a display is given in Figure 2-1 31. Emphasis
is made on the user oriented format and the utilization of application oriented
language leaving necessary conversion and data packing to internal machine
codes.
I
REFERENCE POINT
LATITUDE:
LONGITUDE: •
DATE: Y -
TIME: - -
-,M - -,D
- - Z
Figure 2-13
Example of Format Display
2-51
CASE (A)
CASE (B)
FIGURE 2-14
2-52
2.3.4.3 Programming Aspects
2.3.4.3.1 Programming System Requirements '
To meet the requirements posed by the application characteristics
identified in Section 2-2, both hardware and software considerations must
be given. The extent of the software will, of course, depend upon the
specific features provided by the hardware.
In designing the programming system it is necessary to recognize
the following "internal" operating characteristics:
1) Many short demands - Perhaps the single, most important observa-
tion regarding the operation of multiple, on-line stations is
the large number of short duration service demands made on the
processor. These demands generate individual information entries
that are built up into a complete message. In turn, a number of
messages may be connected to form a complete transaction.
Consider, for example, a simple alphanumeric message entry by a
keyset station. The operator would press a key indicating his
desire for service. The computer would respond with a ready
signal (display) or would present a message (display) requesting
the operator to select a format number. The operator would now
enter a selected code-say 2 characters-and press the 'tnd of message'
key. The selected format would be displayed, and the operator
would enter his data.
The sequence performed by man and machine up to, and including,
the first character of data entered into the selected format is,
using the steps shown in Figure 2-15 as follows:
CO^iSOLE
COnPUTER
AUXILIARV STORAGE
START
1
PRESS FUi;CTION
KEV
COMPUTER PRESENTS
CRT DISPLAY t
LIGiiT COiTROL
Interrupt
I
ENiER D/\YA
INTO OI3f'L;'/
AS KEOUIflcO
INTERRUPT
DETERMINE MESSAGE
S 3 GNAL
SEEK
APPROPRIATE
PROGRA/t
y
INITIATE
PROGRA^t
J
^
SEEK
ASSOCIATED
DISPLAY
}.
TRANSMIT DISPLAY
S LIGHT CONTROL
ACCESS A.XD READ IM
PROGRAM
IF NECESSARY
ACCESS AND READ IN
DISPLA/
FIGURE 2-15
CONSOLE/PROCESSOR SYSTEM OPERATION
I
2-54
I nterpretat i on
Indicates desire for service
Generate response
Receive ready signal/message
Continue if no error, otherwise back to a)
Enter first character of format number code
Accept input
Validate the character
Continue if no error, otherwise back to d)
Enter second character of format number code
Accept input
Validate the character
Continue if no error, otherwise back to g)
"End of message"
Validate the format presented
Continue if no error otherwise back to a)
Enter first character of alphanumeric message
Accept input
Validate the character
The typewriter itself provides an automatic presentation of
inserted data and, therefore, for this example, frees the computer
from those steps 3 marked with an asterisk.
Step
a)
A
1,
2
,3
b)
B
c)
C
d)
D
1,
2
.3^A-
e)
B
f)
C
g)
D
1,
2
,3-A'
h)
B
i)
C
J)
A
1,
2
,3
k)
B
C
m)
D
1,
1,
. 3v^'
n)
B
2-55
The total internal processing time for the above actions may be
about 50 milliseconds for a medium speed computer. In addition, up
to several hundred milliseconds may be required for accessing and
presenting displays if they are needed and auxiliary storage is used.
Since the total elapsed time for the entry of the above example may be
about 10 seconds, it can be seen that very little of the total computer
capacity has been used. If the I/O transfers are buffered, then the
references to auxiliary storage will not be additive with the processing
time. On this basis, a single station would use 50 or 0.5 percent of
]U,UUU
computer processor capacity, and k stations would require 0.5k percent of
capacity. It is interesting to note that for larger and faster computers
the denominator remains constant whereas the numerator decreases. In
fact, for computers of the CP 667 class, this may be only one or two
milliseconds. Hence, it is concluded that a single processor is capable
of servicing many stations and still have large capacity left over for
functions .
2) Time sharing of Processor - Based on the foregoing, there will be a
considerable amount of processor capacity available for tasks other
than the routine servicing required by consoles, A good part, if not
all, will be required to actually execute the functions that may be
initiated by the completed console message or request. In addition,
there will be other computations, related to the total problem or,
for that matter, secondary in nature, that will use up excess capacity.
Because of these interactions and the step-by-step processing associated
with each console, the entire system must be time shared, and more than
likely, multi-programmed.
3) Special Timing - The special nature of the display device interface and
possible electromechanical responses may impose special timing
requirements on the computer program. This depends, of course, upon
how much of the detailed bookkeeping and control is actually committed
to hardware. Consideration must be given to the following representative
i terns :
2-56
a) Refresh - if automatic buffers do not refresh the volatile
CRT's then the computer must re-transmit the information
sufficiently often, say every 20-25 milliseconds.
b) Scanning - it may be necessary to insure the clearing out
of the console output register sufficiently often, say
every 200 milliseconds, if keyboard entry is to continue
at operator pace.
c) Outputs - special devices may require output data that
must meet specified timings. This may be true, for example,
of electromechani ca 1 ly driven devices where start/stop
problems may arise.
The actual implementation of these response requirements will
depend upon the availability of suitable interrupt features,
an active or passive clock, and sufficiently long buffers.
If none of these features are available, detailed programming
may take their place at the expense of possibly affecting
total efficiency. In this regard, an important point is that
timing may not always be critical, and cycles can be skipped
now and then. For example, a CRT image would not suffer from
occasional misses at refreshing, or from a 5-10 millisecond
delay in a cycle. However, while random pertubations will not
adversely affect the viewer, periodic misses will be noticed.
4) Auxi 1 iary Storage - The requirement for rapid response and the
servicing of many stations, each actively spread over a long
period of time relative to the effective computing rate, leads
to the need for rather large capacity auxiliary storage. This
store will hold programs, working displays and the basic data
base. While this discussion points to the need for random access--
bulk storage, such as drums and discs--it is still possible under
appropriate program design, to be reconciled with magnetic tapes.
2-57
2.3.4.3.2 The Display Subsystem Programming System
As discussed above, multi-station, multi-purpose display systems
require random and unscheduled servicing by the computer. Further, the
interactions between man and machine are spread over relatively long
periods of time and are completely asynchronous with respect to each of
the users.
Systems of this type require access to programs, pre-stored display
and data on a random basis if reasonable response times are to be met.
Because of the expected time sharing of the central processor between
multistations for intermittant servicing, it is necessary to maintain
console "history tables" which reflect the transactions that have been
generated by each user to some point in time. In addition, it is necessary
to maintain "position" in a particular procedure since unpredictable time
lapses will occur when human responses are required in fulfilling individual
steps of the procedure.
Effectively, the program must wait (or do something else) whenever
a display is presented to the operator. As the operator enters data
(if required) the computer must momentarily return control and monitor
each entry. Upon completing the entries for a single display, an appropriate
"end of message" will dictate that the next logical step in the procedure
is to be initiated. At the end of the final step, a complete and meaningful
message or direction will have been generated from which the computer will
now determine action independent of the operator. It is thus possible to
continually generate directions and have the computer respond to them on an
overlapping basis.
It is possible to separate the application oriented tasks from those
that are general purpose and apply to most on-line display system applications
and processes . The division is made between the processes required in
generating a message and the actual procedures for executing the action that
may be called. The former concerns the mechanics of handling displays and
composing messages, whereas the latter is concerned with actual file handling,
retrieval, processing, summarizing and formatting. In this disucssion,
attention is restricted to the first aspect, the general purpose processes.
2-58
The objectives for the programming system are as follows:
1) Provide general capability and flexibility so that virtually
all applications can be accommodated.
2) Standardize techniques and procedures so that individual
program segments or subroutines can be shared by as many
functions as possible,
3) Maintain order among contending users for the same files.
4) Service each console as if its operator is the only user
making demands on the processor.
Based on the above discussion, the programming system must include:
1) Display Subsystem Executive Control
This program performs the basic scanning, sequencing, and queue
control for servicing the on-line devices. In addition, it links
to the Master Executive Control which may be supervising the total
processing system.
2) Funct i on Mon i tor
This program maintains the history tables and establishes the
action sequences to be carried out as a function of the keys
that are pressed.
3) Utility Program Package
This is a collection of service routines used primarily by the
Function Monitor and Executive Control. The availability of these
general purpose programs precludes receding of common functions.
4) User Language
This is the language which must be used by the application programmer
in writing his program. The system must provide the programmer
with the ability to express his program in both the symbolic language
of the computer where each command generates one machine instruction
and in higher order languages where each command generates many
machine instructions.
In order to be effective, the higher order language must possess
the following chief attributes:
a) It must be powerful enough to express the application
problem.
b) It must be such that nonprog rammers can use it with a
minimum amount of training,
c) It must be readily expandable so that new commands and
functions can be added.
2-59
The existence of a system such as this implies that application
programmers must conform to certain coding restrictions and procedures so
that all of the possible programs can be accommodated by this approach.
While this may seem a disadvantage, it is, in fact, a saving grace since:
1) It simplifies the programming because of the existing of service
routi nes .
2) It simplifies the implementation of a new application since
the design must fit within the logical framework set forth by
the system.
The importance of the second point cannot be overemphasized. Without
a well-defined organizational and procedural philosophy, the programming
design and implementation of the Individual application can become a major
undertaki ng.
2.3.4.3.2,1 Display Subsystem Executive Control
The real-time requirements associated with on-line displays present
a problem of priority of Interrupt handling and servicing. Hence, it Is
necessary to design an executive system which will be responsive to these
requirements. Such a program will be equipment dependent in the sense
that many hardware/software tradeoffs are possible.
The basic requirement of the display subsystem is the control of a
great number of I/O. This Includes: -
1) Scanning the input lines for messages
2) Refreshing the CRT' s
3) Accessing programs, displays and data from auxiliary memory
k). communicating with other processors that may be in the system
5) Maintaining timing responses for special purpose on-line
display equipment.
"' Initially, it is assumed that the system has a minimum number of
desirable hardware features.
2-60
It has already been stated that typical timing requirements range
from refreshing the CRT within 20-25 ms periods, to scanning of inputs from
the console keyboards every 200 ms. Unless certain hardware features are
available, such as automatic interrupts and I/O buffering, the programs will
have to take these into account.
Assuming no dependence on hardware, the executive program must main-
tain continuous cognizance and control over the I/O. This is done by the
basic control loop shown by the dotted lines in Figure 2-15. Each of the
five indicated functions could potentially generate a processing task as
the cycle is traversed. For example, the tasks associated with scanning
the input message lines is shown in Figure 2-17.
To meet real-time requirements, this loop must be passed at a rate
which will insure return to the task which has the tightest timing constraint
within a specified amount of time. This time will be called the "basic cycle
time". Thus, if the CRT refreshment is the critical task, then the basic
control loop must return to that task within a basic cycle time.
There is also the further implication that the processing requirements
for each of the five identified functions must be completed within a time
which will not compromise the total cycle time.
There are three ways of achieving this:
1) Allow processing to proceed in increments of the basic cycle
time so that temporary return to the cycle is permitted after
each such segment. This leads to difficulties of recursive
entries into the various processing tasks.
2) Spot-place a particular task in more than one position in the
loop. Thus, for example, the "refresh CRT' might be placed in
every other position in the loop if the other functions have a
period which is very much larger than that of the CRT refresh cycle,
3) Permit only a minimum of processing as each of the tasks are
reached and place in a queue those functions not completed. This
queue is then processed during the residual time which is left
over during every cycle. This is shown in Figure 2-16 by the box
which is part of the loop indicated by the heavy lines. It is,
of course, necessary that the residual be non-zero enough of
the time if any processing is to occur.
2-61
Start
1
Scan Input
Lines
Refresh
CRT
Perform I/O
References and
Transfers
Maintain timing
for special
purpose devices
J i
Commun icate
with other
processors
Task Queue
Processing
Figure 2-16
Basic Executive Control Loop
2-62
Scan input
1 ines
no
message
t
message
Decode signal
source
Decode signa
type
Continue basic
control loop
Service
Response
Val idate
step
Function
Response
Place in queue
for Function
Moni tor
Routine
in Core?
no
Place I/O
Request in
I/O Queue
yes
Execute program
or
place in queue
Figure 2-17
Tasks Associated with Scanning the Input Message Lines
2-63
A particular system will include any number of these possibilities
depending on the details of the system interface and hardware characteristics.
The example shown in the figure employs the last of the above alternatives.
If one has m consoles in the system and each one is generating input
at the nominal 60 wpm rate, then m consoles would require servicing every
200 milliseconds. Based on experience, the typical processing time per
character (command) entry is one millisecond for computers of the 12 micro-
second memory class. Hence, if m is 20, then 20 milliseconds out of every
200 (or 10% of the available processor time) is spent in console servicing
i f the entry rate is sustained.
A more realistic analysis of processor support to displays is given
in Figure 2-ia. Here it is shown that if messages can be entered in 30 seconds
a single console will require a total of 1.0 seconds of processing and 1.6
seconds of I/O time assuming a 12/'s. computer memory and the availability
of a fast disc system for auxiliary storage. Azimuth m=20 consoles would
require 20 seconds of the processor's time and 32 seconds for I/O. The
latter Is, of course, impossible in a 30 second period unless there are
multiple buffered channels in the system. It can be seen that a multiple
console configuration can saturate the computer capacity so that while all
of the consoles may be serviced, the other processing tasks are not satisfied.
Based on tables such as shown In the figure, a proper balance of number of
stations, speed of processor and total processing tasks Is achieved as a
result of a system analysis.
Total Processing Time
Rate of
Average
Number
Occurrence
Processing
Per
Time
Message
Comput i ng
I/O (Assumes
d i sc)
Alphanumeric Entry
200 ms
1 ms
150
150 ms
Display Change
4 sec
250 ms
8
400 ms
1600 ms
Function Key
6 sec
150 ms
5
250 ms
Complete Message
30 sec
200 ms
1
200 ms
Total
1000 ms
500 ms
Figure 2-18
Processor Servicing Required in Support of Console
Message Entry
(exclusive of refreshing)
I
2-65
The following comments are presented concerning the implementation
of the executive control with respect to the presence or absence of the
indicated hardware features:
1) If neither external interrupt nor a real time clock are
available, then the tasks associated with each of the
control loop functions and all other calculations must
be programmed in segments so that each segment will permit
return to the control loop and maintain the timing.
2) If a clock is available then the executive can preset it
at the beginning of each cycle so that it will interrupt
the processing of the queue at the proper time,
3) If external interrupts are available then the function
of the basic control loop has been absorbed by the
hardware and no executive function is needed. Consoles are
then serviced on demand.
2.3.4.3.2.2 Function Monitor
The function monitor is a specially designed program to facilitate
the responses to a special set of keys on the console. Although not all
consoles have a set of keys of this type, it is deemed necessary that a
truly general purpose console will have such a set. They are characterized
by the fact that their labels and also their identifying codes can be changed
at will by the operator.
The process of entering information into the computer for the purpose
of making a request has been discussed in detail earlier. It is primarily
to ease this process that the function monitor is designed. Knowing that
different applications will require different displays and different
sequences of presentation, it is apropos to design a scheme which is not
application oriented and is professional programmer independent so that
the user can design his own data entry scheme and query language.
2-65
The function monitor is an interpretive program which operates
on a very special language useful in display manipulation. When one
of the special keys mentioned above is pressed, the executive control
recognizes this and passes control to the function monitor. There,
the specific key is identified and an associated table of instructions
in the special display language is executed i nterpret i vely . It is the
ease with which a user can modify this table of instructions which makes
the function monitor so valuable. To illustrate the capability of the
display language some of the possible instructions are:
1) Turn the specified console lights on (off) - the lights
are specified in parameter words following the Instruction.
2) Display the following characters on the CRT - the characters
along with their location coordinates are listed following
the instruction.
3) Locate a display in auxiliary storage - the identification of
the display follows the instruction.
4) Clear a specified buffer - the buffer area may be either
pre-established or specified In the words following the
i nstruct ion.
5) Enter the specified characters In the buffer - the characters
are listed following the instruction,
6) Process the "list" display - special codes (specified by the
query language) are extracted from the list display as
dictated by the selections of the operator and are placed
in the buffer.
7) Process the "format" display - the parameters entered
by the operator are extracted from the format display
and stored In the buffer.
2-67
A more sophisticated language can easily be designed to cover more
applications. The above language, however, is completely adequate along
with its function monitor to service the kinds of retrieval requests set
forth as examples in Section 2.3,2.
2.3.4.3.2.3 Utility Programs
Utility or service programs extend the hardware in a general way so
that certain functions become available to the application programmer without
his concern for programming. This software is primarily concerned with
facilitating the entry of alphanumeric information onto the CRT in an
expeditious manner. Also included are useful functions for data handling
and in the control of displays.
In some instances the recommended features described below may be
part of the hardware, thereby precluding a need for the programming.
1) Marl<^r Routines
The marker is a special symbol which is used to indicate
current writing position on the CRT. The following control
keys are defined for manipulating this marker:
a) Marker Enable - This key causes the marker to appear at
some fixed location on the CRT. This position could be,
for example, the (1,1) character location. As alphanumeric
characters are entered, the marker is displaced one character
position to the right, the newly entered character taking
its place. The marker moves from the end of one row to the
beginning of the next and upon reading the lowest right hand
position, it will return to the (1,1) position. A character
that is dislocated by the marker will be replaced when the
marker is moved again, unless a new character has been
entered.
b) Marker Disable - This key removes the marker from the CRT.
c) Marker Backspace - This key causes the marker to move one
position left, or to the end of a previous line, if now at
the beginning of a line.
2-68
d) Marker Up - Depressing this key causes the marker to move
to a position in ihe preceding line wiiich is directly above
its current position. If the current position is in the first
line, the marker is moved to a position in the last line ver-
tically below its position in the first line.
e) Marker Down - This key causes the marker to move exactly
opposite to the motion described in "Marker Up."
f) Marker Left - Depressing this key causes the marker to move
in positions to the left in the same line. The marker moves
"end around" from the first to the last position of a particular
line. If n=l , then this key is identical to the backspace key
except that the latter is not restricted to a specific line.
g) Marker Right - This key causes the marker to move exactly
opposite to the motion described in "Marker Left" except that
the number of positions moved is n'. A relationship should
exist between n and n' such that one of them is equal to one
and the other is some small integer greater than or equal to
one, A recommended system Is n=5 and n'=l.
h) Advance Marker - This key is used in conjunction with the
format display, i.e., a display in which the operator enters
A/N data into various labeled slots. Depressing this key
causes the marker to be moved from its current position in
some slot to the first position of the next slot. If the
current position is at the last slot, the marker is moved
to the first position of the first slot.
I) Accept I tem - This key affects the marker only with respect to
list displays. The depressing of this key will move the marker
along the first column, from one row to the next, replacing
the marker by "X", indicating that a particular item was selected
j) Reject Item - This key affects the marker only with respect
to list displays. The depressing of this key will move the
marker along the first column, from one row to the next,
replacing the marker by "space". This feature is used to
reject a previously accepted item.
2) Display Control Keys
The operation of the CRT display is aided by the availability
of the following keys. For convenience, a distinction is made
with respect to the CRT display which is viewed by the operator
and the CRT display image, (or just image) which is the computer
stored analog of the CRT display.
2-69
a) Display On - This key causes the CRT display image Lo be
presented on the CRT.
b) Display Off - This key removes the CRT display leaving the
Image in a passive state.
c) Clear Display - This key causes the CRT image to be completely
cleared except for the marker which, if on, is restored to
its origin.
d) End of Message (EOM) - This key is used in conjunction vyi th
data entry to indicate to the processor that a message has
been completed. It serves as an interrupt which signals
the computer to act on the CRT data,
e) Data Insert - This key is used in order to insert a set of
alphanumeric data on the CRT between two consecutive characters.
The marker is first positioned to the leftmost of the two
characters. Then the Data Insert key is piessed and new data
is entered appearing as it is generated and causing all of
the data to the right and down to be shifted by one position.
Exit from this mode is made by pressing the EOM key.
f) Data Delete - This key is used to delete a set of continuous
alphanumeric data on the CRT, followed by a closing up of the
display. The marker is first positioned at one end of the set,
the Data Delete key is pressed and then the marker is set at
the other extreme. Pressing of the EOM key causes the desired
action and exits from this mode.
g) Sequence Di splay - This key is used to call for the next part of
a multi-part display should the size of the CRT prohibit the
display of the entire message at one time.
h) Display to Printer - This key generates a hard copy version
of the CRT display on an associated typewriter or line printer,
whichever is available.
i) Moni tor Di splay - This key permits the selection of any other CRT
associated with another console for purposes of monitoring that
console's activity.
j) Save Display - This key interchanges the CRT image with the
contents of an alternate location. Thus, effectively it
permits saving Information for future reference purposes.
Typically, after pressing this key one will also press Clear
Display if one is disinterested in the display brought forth
from thealternate image location.
2-70
3) System Control
In this section representative functions are identified and
assigned to the Control Keys, In a particular system more
descriptive and extensive keys may actually be called for.
a) Display Message - This key permits interruption of the current
CRT display for purposes of viewing the message which is being
held by the computer for the operator. The availability of a
message is indicated by a status light (see below). Return to
the current procedure is by pressing of the Display On key.
b) Display Q.ueue - This key causes the internal tasks queue
(if there is one) to be displayed. Shown are priority ordering
and status. The operator is now able to modify this queue by
manipulating the CRT display and using the Modify Queue key
(see below) .
c) Modify Q.ueue - This key can only be operated after the Display
Q.ueue key was pressed. It causes the CRT display to be sent
to the processor where the queue is then modified.
d) Change Procedure - This key provides a display which permits
the operator to modify, select or cease system operation.
Typically, this feature is an overall control procedure
which should be assigned to only one of the on-line stations.
e) System Breakpoint - This key is essentially an external interrupt
which performs two functions. The first is to save-store system
status for rollback purposes in case of hardware failures. The
second is for modifying the system configuration or operating
procedure.
4) Status Indicators
Status indicators reflect the composition of the configuration,
intermodule communication situation, internal machine control situation
and system operating modes.
^) Power On - Indicates whether console is in operating mode.
b) Processor not Communicating - Indicates if the communication
between console andprocessor has lapsed more than some pre-
established period of time (say 500 milliseconds).
2-71
c) Queue Fu 11 - Indicates that the Internal task queue is
full, and that no further inquiries can be made of the system.
d) Message Ready - Indicates that a message has been generated
by the processor for the operator. The operator can select
this message on the above-mentioned Display Message key, which,
when selected, turns this indicator off unless a second message
is a Iso present.
e) Operating Mode - Indicates which mode is currently in operation.
An indicator is dedicated to each operating mode identified by
the system.
f) Conf igurati on - Indicates which peripherals are on-line with the
system. An indicator is dedicated to each of the relevant
devices. This indicator is useful as a means of assigning perip-
herals to different consoles, It is used to display legal or
illegal connections for any one console.
5) Error Indicators
The following alarm indicators are indicative of the signals that
are useful to the operator. These indicators have an associated
button with which the operator can cause a "reset" action to take
place and attempt the procedure once more. The indicators should
be placed in an obvious position so that the operator will be
cognizant of alarms. One procedure is to cause the indicator to
blink on and off at an appropriate rate, say twice per second.
s) Par i ty - Indicates parity error in transmission from, or to,
the console.
b) Keyboard Locked - indicates that illegal use was made of the
keyboard, such as pressing two keys within a disallowed time
i nterva 1 .
c) Data Entry - Indicates that some rule regarding data entry on
the CRT was violated.
d) Procedure - Indicates violation of order regarding the use of
the function keys.
e) Control - Indicates violation of rules regarding the use of a
control key.
2-72
2.3.5 System Design Steps and Considerations
2.3.5.1 Identification of Evaluation Parameters
The determination of "best" display is a function of system balance
where cost, computer programming, and demands on the computer must be
measured for the application.
While the first of these is evident and simple - i.e., a dollar cost
for the display and all interface boxes and cables - the second is more
elusive, while the third is often a neglected consideration.
2.3.5.1.1 Display Hardware Costs
Display hardware costs are not only measured by the cost of the
particular keyboard and CRT unit but must also include the black boxes
and cables which connect the device to the processor. Total display sub-
system costs are also a function of number of units. Since displays are
often custom designed to each user's specifications, single unit purchases
are usually more expensive than buying them in lots of five or more. Also,
in many cases, parts of the hardware can be time shared, and the unit price
decreases as the number of units increase.
It is desirable to consider alternatives in system configurations since
cost is related. There are several methods for tieing the on-line device
to the computer. The appropriate method for each application must be
determined during the system design when the system equipment is being
specif fed.
2.3.5.1.2 Computer Programming Requirements
The use of on-line communication devices places software requirements
upon the total system. The extent of the software which is developed will
depend upon the specific features provided by the hardware. For example,
the programming developed for a specific display console connected to a
CDC 150 computer system Is given In Figure 2-19.
CDC 160 PROGRAMS
PROGRAMS
MASTER CONTROL PROGRAM
150
A/N SUBROUTINE
125
A/N LEGALITY
15
CHARACTER TO CRT DISPLAY
32
COMPUTER MARKER POSITION
20
MARKER KEYS PROGRAM
131
ENTER/CANCEL
20
REGENERATE CRT
14
ERROR LIGHTS SUBROUTINE
17
TAPE SEARCH ROUTINE
246
CRT DISPLAY
20
DISPLAY REQUEST QUEUE
16
CRT DISPLAY TO 1604
46
CHANGE REQUEST QUEUE
19
MODE/MODEL CHANGE
186
STOP MODEL
3
NO DISPLAY
37
CLEAR DISPLAY
29
DUPLICATE DATA BASE
3
1604 ON-LINE
18
1604 MESSAGE READY
104
INSERT ROUTINE^^^
155
DELETE ROUTINE^^^
96
OVERLAY INTERPRETATION
556
PROGRAM
SEQUENCE DISPLAY
86
BUFFERS
CONSTANTS
64
CRT DISPLAY IMAGE
448
CONSOLE LIGHTS
5
OVERLAY TABLE
512
SUPPLEMENTARY RECORDS
256
TOTAL WORDS
3429
o
o
o
o
o
<u
c
c
O) o
— o
0)
u.
13
cn
0)
'o
c
o
o
>^
05
o
E
cn
O
2-74
2.3.5.1.3 Demands on the Computer
To L^nalyze the demands on the computer system by on-line consoles
it is necessary to define a problem mix and the detailed types of operations
that will be employed in the execution of the task. The typical problerii
studied is the composition of a query to the data processing system.
To carry out this job, the operator will:
a) Depress function key 1
b) Get positive response that key 1 was depressed
c) Computer presents a display on the CRT
d) Enter data into display
e) Visually validate inserted data
f) Make corrections to inserted data if necessary
g) Signal end of entry
While this sequence of user/operator events is underway, the computer
is engaged in a number of actions paced by the speed of the console operator.
The conclusions are to be drawn. First the tieing of on-line displays
to a computer will require dedication of memory.
The second conclusion concerns the amount of computer time actually
used by the displays for display activity independent of retrieval, formatting,
and presentation.
The third conclusion concerns the potential traffic problem which
multiple consoles may cause with respect to the data channel to which they
are connected and with respect to the I/O transfers required between auxiliary
storage and processor and in the processor itself.
Using the data presented earlier as a basis, we can obtain an upper limit
on waiting for a multiple console system. We assume a worst case model where
the total processor and I/O time of 1.5 seconds is lumped together as the service
time . Using the theory associated with Poisson processes, we can estimate
the waiting time form knowing the service factor . This number is the ratio
of service time to total elapsed time between requests and is 1.5/30 or 0.05
for the problem at hand. The results of the traffic analysis are given in
Figure 2-20 where a service factor of 0.03 is also added. The latter figure
leads to a model which assumes buffering and better organization of the
processing tasks.
Probability of n Consoles
requ i r i ng serv ice
Number
of
Serv ice
Serv ice
Number
of
Conso
es
Factor
Factor
Consol
es
Wa i t i ng
= .05
= .03
.538
.712
1
.269
.214
2
1
. 121
.057
3
2
.049
.013
4
3
.017
.003
5
4
.004
.00 1
6
5
.001
.000
7
6
Cumulative Waiting Probability
Service Service
Factor Factor
= .05 = .03
121
170
i87
192
193
057
70
073
074
000
.000
93
.074
Figure 2-2o
Probability of n Consoles of Ten Requiring Service at
the Same Time
i
en
2-76
The results shown that a waiting time will exist 19/ of the time for
the 0.05 service factor. Since the service time is 1.5 seconds, the average
wait on the waiting line will be sliyhtly under 3 seconds. For the second
mode 1 , a wa i t i ng 1 i ne w i 1 1 ex i s t abou t 7 . 4'/, of the t i me . The ave rage wa i t on
the waiting line will be approximately one service time of 1.5 seconds.
These results are sufficiently favorable and tolerable that in a practical
sense the console operations will experience no appreciable waiting, especially
when it is realized that these statistical estimates reflect a pessimistic
mode 1 .
2-77
2.4 INPUT/OUTPUT TECHNOLOGY
2,4.1 C 1 Jss i f i cj t i on of Input-Output Technology
Input-output technology deals with the techniques which a computing
bystem uses to communicate with the outside world. Functionally, there are
two different classes of subsystems in the outside world with which the
computer must exchange usable information. The first subsystem is the
human, who communicates in a wide variety of non-exact languages that require
elaborate interpretation. The second class of subsystem is the non-human
or machine, which uses a relatively smaller number of languages, all of
which are exact and defined. These two problem domains are quite divergent.
The relations of man and machine in a typical military information system
are shown in Figure 2-21 «
2.4.1.1 The Man-Machine Interface
Although a machine (and its attachments) is quite versatile in its
ability to sense a wide variety of inputs, e.g, visual, sound, pressure,
radiation, etc, man is capable of producing only two outputs which are
relatively controllable. These are sound and pressure or motion.
2.4. 1.1.1 Sound
The human is able to produce a greater bandwidth of information
vocally than in any other manner. This information is produced with
built-in identification characteristics such that two people may be talking
at the same time and yet their conversations may be distinguished from each
other. Unfortunately, the associative characteristics of human thought are
such that it is difficult for the same person to express himself in exactly
the same terms tvvice in succession, and it is nearly impossible for two
different people to express the same thought in the same manner.
MAN
/ /
/ / ./
/
/ / /
MACHINE.
7—1 — 7 — 7 — 7 ^^ '^ — ^
/ / / / /' -' / . Receives/and Displays Information
V / / / / / / / / ■ / ,- cJU
/
Analyzes Status of Forces
77~"
/ / /■ /
/
/' / / / / Updates S i tuat ion' Di splay / /
/ / / /' / / /' ,- c . / / ■-- ' / ^
Evaluates Military Situation
/ ■ / / / / / / / / / /
/ / / / / /■ /
J. / / / .. L c
Responds to Data Requests
/ /
Identifies Courses of Action
/
/ /
-> — 7 7 y
Computes Hypothet i ca 1 'Effects
/' / -■''
^ .
/ / / / ■' /
Makes Decision
/
7 — 7—7 7 — -7 — -? 7 — 7 ~
ommunicates and Records Commands
/
FIGURE 2-21
Relations of Man and Machine in a Typical Military Information System
00
2-79
Although sound input transducers for computer usage are relatively
inexpensive, vocal human interface is seriously hampered by the lack of
an interpretive concept to allow the machine to understand the wide
varieties of expression that a human may produce, even when vocalizing
a concept held constant,
2,4. 1 , 1.2 Pressure or Motion
The only alternative means of man-machine communication is the use
of pressure or motion. Here, a human is quite inefficient, being able
at a peak to produce only about three hundred controlled, distinguishable
yes-no motions per second. At this rate, the motion must be of a reflex
nature and the data involved must be preconceived and prerecorded. The
10-key adding machine operator can copy data at a peak rate of about 20
numeric characters per second, when selecting these characters from a
total of ten possibilities. A good typist can select from about 50
characters at the rate of ten per second.
It is evident that although the action speed increases as the choice
is reduced, the total bandwidth of information that can be transmitted
increases as the action rate is reduced; thus, even greater information
flow can be created in a situation in which a computer presents to a human
a number of complex alternatives and the human makes a selection of the
alternatives he wishes. Here, although there Is a yes-no decision made by
the human, the information content of this yes-no decision is quite great
because of the human's preprocessing of a volume of data to make the
dec i s ion .
- Numbers refer to references listed at the end of each subsection.
2-80
2.4.1.1.3 Human Language Interface
There ex is Lb a category of information transfer in which the information
to be transferred is machine recorded in a human language. To further
process this information, it is necessary for the machine to be able to
read the human language even though the data itself is not being, ut this
point in time, originated by the action of a human. Typical of such
human language interface machines would be character recognition
equipment. The problems in the design of such equipment are similar
in nature to those that occur in the design of equipment where the data
2
is actually originating with the human. All the vagueness and lack
of exactitude of human language exist within the data and a rather
3
sophisticated means of interpretation is required. One might think,
however, in just reading and transferring, that this can be done by
blind rote if the meaning does not have to be deciphered. The parallel
is not really exact since the data is only being transferred as over a
communication link, and a true man-machine interface does not exist.
Telephone lines certainly deal with the human language but they need a
human at each end. Whenever the data has to be entered Into a machine
for the machine to operate on the data, the Interface exists and the
problem of data interpretation has to be solved,
2.4.1.2 Machine-Man Interfaces
Sensitivity of a man Is such that he Is quite limited In the number
of techniques b which he may receive a reasonable quantity of meaningful
information. There are in fact, only two channels available with useable,
effective bandwidths. These are visual and auditory. Through both
channels the man is able to sense a wide disparity of Information, select
that which Is of interest to him, reject all superfluous information, and
fill missing gaps from context or redundancy.
2-81
The contpuLet', on the other hdnd, is noL a pdilicularly good generator
of audio visual information despite the work at synthesizing speech using
canned phrases or phonenies and the sending of audio codes v^hich might be
intei'esting in some application. Although computers cari generate complex
displays, I heir ability to produce and display visual informaticjn in no
way approaches that of the human. This is probable due to Its own limited
language structure and the lack of variety In ways in which a computer can
express I tse 1 f .
2.4.1.2.1 Visual Interface
As the human is an excellent classifier, sorter and filterer for
information, he is capable of accepting a very wide band of visual Input,
taking cognizance of those items of interest to him and Ignoring all other
I terns until they reach a status that calls them to his attention or until
he reaches a status that calls them up. The bandwidth of Information which
he is able to accept visually is related to the language in which it is
presented and the human's facility to handle that language.
Typical of the languages in which a humancan accept information
V I sua 1 ly are:
1) The various printed and symbolic representations of
spoken languages,
2) Non-spoken symbolic representation languages such as mathematical
formulae and chemical formulae,
3) Geometric forms, diagrams and other forms of special
relation intelligence,
4) Miscellaneous visual differences such as color and motion.
As there is a great deal of difference In the technology necessary to
generate these different forms of visual presentation, and as the different
forms are used for quite widely divergent functions, visual interface will
be discussed in two separate parts of the technology study. The printed
and symbolic forms of spoken language, some type of graphic and geometric
communications and display of formula, when presented on a permanent
document, will be considered under input-output equipment. All forms of
2-82
visual communication, when not created for record purposes, will be
considered as display equipment. In present day connotation, display
equipment implies a degree of real-time response or rapport betvveen the
human and the computer,
2.\. \ .1.1 Auditory Interface
Auditory interface between a computer and a human can exist In two
different ways. The human can be trained to recognize some form of
auditory output of the machine code. Such an artificial system could
be devised to allow the machine to generate Morse Code." Changes in
repetition rate of a signal may shift the frequency of a tone, or
"operate" commands can ring a bell. All such forms of sound discrimination
provide a very narrow bandwidth of communication between the machine and
the human.
An alternative means of sound communication between the machine and
the human is to allow the computer to generate, or select from storage,
an appropriate series of phonetics, words or phrases and assemble them
into a meaningful spoken sentence.
A human receptor is quite capable of tolerating and filtering out
noise and other unmeanlngful trivia and, where necessary, filling in
missing gaps from context. Even a relatively crude human vocal
simulation can transmit meaningful information between the machine and
the human. The human can receive a bandwidth up to 300 words per minute
and at relatively low noise levels.
This has, In fact, been done to allow a program to send over Its
audio console monitor, the path which a complex program Is taking during its
cycling In non-real time (or free-time).
2-83
2.4.1.3 Mach i ne- to-Mach i ne Communication
The problem of communicating from one machine to another is quite
different from that of communicating from a man to a machine or a machine
to a man. The difference is that the human has already been designed and
his limitations must be accepted, whereas a machine may be designed to do
a specific job. The result is that a machine may use any media for
communication with any other machine and the two machines may jointly use
any conceivable coding system. Mach i ne- to-mach i ne communication, therefore,
is essentially a question of coded energy transfer. The efficiency of
mach i ne- to-mach i ne communication depends upon the efficiency of energy
transfer of the media selected and the true data content of the coding
system used. The reliability of the communication will depend upon
the redundancy of the code used and the amount of noise or interference
4
which occurs during the communication.
Two different sets of criteria may be used in the analysis of machine-
to-machine communication. These are:
1) The function of mach I ne- to-mach i ne communication
2) The technique of mach i ne- to-machi ne communication
2.4. 1.3. 1 Mach i ne- to-Mach i ne Communication Functions
It is obvious that the prime function of mach I ne- to-machi ne communica-
tion is the transfer of data; however, this data may be transmitted to or
from a machine to provide data which the other machine will work upon, or
it may be transferred to the machine to control the machine.
Such control data, unlike information data, frequently requires the
transmission of power to drive a unit, (e.g. close a relay, close a valve)
or it requires the transmission of an analog, (e.g. a change in voltage, a
change is pressure, etc.). Although not frequently recognized by the
digital engineer when he lifts the level of a line or pulses a line with
an on or an off pulse, he creates an electrical analog of the opening and
closing of a switch which in turn opens or closes a second switch. Such
control information may be considered analog unless it is transmitted
through a series of digitally coded pulses.
2-84
The most efficient means of transferring information from one point
to another is usually the use of digitally coded data. Efficiency is
gained by allowing more than one type of data to be transferred over a
single line. Where a line exists between two points, any combination of
pulses may be transmitted over this line. The data transmitted over this
single line may be used by a multiplicity of different types of equipment
all attached to the common line but each equipment capable of listening
for its own coded "call signal" and decoding the data that follows.
2.4.1.3,2 Techniques of Mach I ne- to-Mach i ne Communication
As with man- to-mach i ne and machine-to-man communication, the two
critical factors involved with machi ne- to-machi ne communication are the form
in which the data to be transmitted exists and the efficiency of compatible
transmitting media. There are essentially three classes of data trans-
mitting media available to the computer designer. These are:
1) Mechanical transmitting media including pressure, movement, sound
2) Electrical conductivity
3) Electromagnetic radiation including heat, light, and radio waves
Within each of these three major categories, there are many sub-categories
which could receive consideration for data transmission in some special
app 1 I cat i on .
2.4.1.3.2.1 Pressure, Movement and Sound
This study is concerned with communication of data from one machine
to another rather than a broadcast of data for general receipt. Pressure,
movement and sound media must be considered as directed or ducted devices
when the data is transmitted from one machine to another. in general, the
frequency response of pressure, movement and sound systems is much lower
than that of electrical conductors. In addition, the propagation rate of
sound is very much lower than that of electricity, resulting in undue delays
where a feedback system Is involved. The one great advantage of pressure,
movement and sound systems is that they have the inherent ability to transmit
relatively large amounts of power from one machine to another and have,
therefore, found application in the process control field. In some cases.
2-85
it has proved to be economical to use these media as a form of data
communication by virtue of the fact that the data already existed as a
pressure or movement and would be used as a pressure or movement by the
rece i v i ng mach i ne.
2.4.1.3.2.2 Electrical Conductivity
At this time, most of the technology used in the design of digital
computing equipment utilizes the controlled flow of electrical energy along
wires. As a result, all input must be converted Into electrical pulses and
all output exists as electrical pulses unless otherwise converted. There
appears little likelihood that there will be any change in this situation within
the next 20 years. If anything, better transducers, microminiaturization of
equipment, larger production volumes, and Improved production techniques
probably will produce an even more entrenched position for the electronics
industry. The communication media required for electrical conductivity (a
length of wire) Is Inexpensive. It has a very high propagation rate and
a wide bandwidth. In most cases, no transducers are involved since the
information both exists, and Is required, in electrical form. Without
doubt, electrical conductivity will continue to be the major means of
mach I ne- to-mach i ne communication in the 1970-1980 period.
2.4.1.3.2.3 Electromagnetic Phenomena
Electromagnetic phenomena including radiated heat, light, and radiowaves,
have a propagation rate roughly equal to that of electricity in wire. They
possess two drawbacks In their application to mach I ne- to-mach I ne communication:
1) They do not readily lend themselves to "ducting" and, therefore,
dissipate large amounts of energy In the process of transmission
and allow the receiver to pick up unwanted energy from other sources
requiring that the unwanted energy must be filtered out.
2) The transducers required to create a carrier, modulate it, receive
It, radiate It, demodulate It and amplify it are relatively less
reliable than equipment designed to transmit through a fixed
conductor, and their use must be justified and more expensive.
Electromagnetic radiation, does, however, have one advantage as a media
for mach I ne- to-mach I ne communication. It allows the rapid transmission of a
2-86
wide band of data from one point to another when the two poiiits are mobile
in relationship to each other, thereby allowing mach i ne- to-mach i ne
communication when one or more pieces of the system is in motion relative
to the other pieces. It also allows fast set up of equipment under field
conditions since no interconnections are required. For these advantages,
electromagnetic radiation pays a heavy penalty in cost, complexity, and
unrel i abi 1 i ty .
2.4.2 Sources of Information
The following sources of information are the ones that have been
dealt with to date. It is anticipated that as this study continues, there
will be additions made to both the people and companies contacted and the
1 i terature used .
2.4.2.1 SOURCES OF INFORMATION - PEOPLE AND COMPANIES
Analex Corp.
Mr. John Simms
Disc Files and Printers
Army Electronic Research £- Development Group, Computer Division
Ft. Monmouth, New Jersey
Input-Output Equipment
Mr. Burkhart, 53-51241
Mr. McGee, 53-51446
B r i dge, I nc.
Phi ladelphia
Mr. Lou bauerwin
Card Readers and Card Punchers
Bryant Computer Products
Di sc Fi les
Control Data Corporation
St . Pau 1 , Mi nnesota
Mr, Bob Windsor
Peripheral Equipment Dept. Computer Division
Mr. D. E. Lund Strom
Product Planning Peripheral Equipment Division
Cook Electric
Incremental Magnetic Tape Recorder
Data Equipment Co.
Tust i n , Ca 1 i f orni a
Digital Plotters Graphical Input Methods
Mr. Raymond Davis
2-87
Di g i da ta Corp.
4908-45 Ave.
Hyattsville, Maryland
Phone: 301-277-9397
Incremental Magnetic Tape Recorder
Digital Equipment Corp.
Maynard, Massachusetts
Ana 1 og- to-Di g i ta 1 and Di g i ta 1- to- Ana log Equipment
General Dynamics Electronics
San Diego, California
Mr, James Redman
Manager, Gov't. Requirements
Mr. R. Glaeser
Manager, Requirements Research Printers
General Kinetics
Variable Speed Magnetic Tape Reader
Honeywel 1 Corp.
Boston, Massachusetts
Mr. V i nee Porter and Mr. Dave Bernard
Input-Output Equipment
Phi Ico Corp.
Mr. Gordon Gibbs
Character Recognition
Potter Instrument Co.
Magnetic Tape Transports, Printers
Radio Corporation of America Laboratories
Princeton, New Jersey
Dr. Jan Rajchman
Solid State Magnetic Tape Unit
Soreban Engineering, Inc.
Melbourne, Florida
Royal McBee Industrial Products Division
Paper Tape Equipment
Sylvania Corp.
Newton, Massachusetts
Mr. D. Lilly
Read-only magnetic Cards
2-88
Sylvai.ia Corp.
Newton, Massachusetts
Mr. R. D. MacNaughton
Mr. R. A. Barbary
Militarized Magnetic Tape Transport
Ta 1 ly Corp.
Seattle, Washington
Punched Paper Tape Equipment
Uptime Corp.
Punched Card Equipment
Wyle Labs
Mr. E. Gamson
Input-Output Keyboard £- Display Unit
2.4.2.2 Sources of Information - Literature
A list of references pertinent to the study of input-output technology
is given in the Bibliography. Some of the material presented in subsequent
parts of this section has been extracted from these references. During the
remainder of this study, the more pertinent and Important of these references
will be studied in more detail and new references reflecting materials pub-
lished or discovered subsequent to the preparation of this Bibliography
will be i ncluded .
2-89
2.4.3 Input-Output Technology Characteristics Required for ANTACCS
and Their Application in the Naval Environment .
This Section is largely a requirement function and, therefore, it
depends heavily upon information to be obtained from the study being
performed by Booze Allen Applied Research, inc. Work on this section
has, therefore, been postponed until better information is available as
to the requirement of future Naval Tactical Data Systems and the environment
within which they are expected to operate. It is anticipated that the
requirement study will furnish information as to the data flow, the sources
of data, and the form in which the data occurs or is required. Such
Information will allow us to obtain a better perspective of the Input-
output technologies In relationship to the Naval environment.
2.4.4 Current Status Review
The purpose of this section is to provide a review of current
technology in the input-output area. It is intended that the technology
covered be that technology embodied In currently existing equipment and
modifications of current practice.
2,4.4,1 Man-machine Interface
Currently, the man-machine interface has not been heavily exploited.
In most computers, there is a man-machine interface in the form of an
alphanumeric keyboard and some function switches, both of which are
usually used only in conjunction with program debuging and machine
operations. Other than this, the man-machine interface seems to be
limited to the command and control area where the human must be interfaced
as a part of an open loop control system.
2.4.4.1.1 Sound
There is no known present equipment where a human generated sound
is used as a computer input. Laboratory work is being done in this area
and will be discussed in a later section. However, it is not possible
for the systems designer to specify a human generated sound input for
current or near current delivery.
2-90
2.4.4.1.2 Pressure or Motion
There are three classes of pressure or motion devices that are currently
available. These are keyboards, function switches, and position indicators.
Theoretically, any of these may be used either off-line or on-line. in
practice, certain types of information such as instructions in human language,
are stored up for later use while other types of information, such as function
selection, are used as a part of a feedback loop.
1) Keyboards
Keyboards are designed primarily to enter symbolic representation
of human spoken languages. These symbols or letters are usually sup-
plemented with other non-spoken symbology. Keyboards may be numeric,
alphabetic, symbolic or any combination thereof; they can be designed
to meet any need.
Although many non-standard keyboards ar& designed for special purposes,
there are three standard keyboards that are accepted in this country:
The alphanumeric or typewriter keyboard, the numeric ten key keyboard,
and the numeric bank or columnar keyboard. There are many variations
within each of these standards. However, there is enough standardization to
allow the training of personnel in their operation.
Alphanumeric keyboards are designed to operate at a peak repetition
rate for a single character of ten or fifteen times per second. As
must alphanumeric keyboards are not interlocked to prevent the
simultaneous depression of characters, it is possible to operate
such keyboards at speeds up to 20 characters per second providing
that the same character is not repeated in sequence. Typical
operator rates are about five characters per second when copying
from legible data.
2-91
Ten-key numeric keyboards are designed to be operated by one hand
using the middle three fingers for the digits 1 through 9 and the
thumb for zero. There is no horizontal movement of the hands required
in such a keyboard and it is possible to obtain fast operator
speed. A trained operator can produce output at the rate of ten to
twenty characters per second for reasonably long periods of time.
The bank or columnar keyboard provides a column for each digit
position. Each column contains all of the digits which may be inserted
in that position, usually 1 through 9. This keyboard Is a type of
forced entry device In that the format is produced In all zeroes except
where digits have been added. Further, It is impossible to enter an
unacceptable digit in the wrong column. This is avoided by omitting
unacceptable digits from the column for that digit position.
The bank keyboard Is frequently used In applications where close control
over the entry Is required. Unlike the ten key keyboard, the bank
keyboard Is operated by hand movement rather than finger movement and
requires the entry of only non-zero digits. A trained operator will
enter more than one digit at a time in the bank keyboard by pre-
positioning her fingers prior to moving her hand to the keyboard
and depressing it. In this manner, the number 871,532,000 would be
entered in two movements or key depressions. The 8, 7 and 5 (digits
1, 2 and 4) would first be entered as a single movement by the
operator. The operator would then lower her hand on the keyboard
and enter the digits 1, 3 and 2 (digit positions 3, 5 and 6). The
last three zeroes would not be entered as they are already standing in
the mach I ne.
This type of keyboard is particularly desirable where dealing with large
numbers that include a number of following zeroes and in applications
where a format control Is required. Each column is usually Inter-
locked so that not more than one number can be entered. It is,
possible to make an error by depressing the wrong digit key. However,
it is difficult to make an error in the magnitude of the number such
2-92
as might be made with a ten key machine by omitting the last zero
or by inserting an extra zero accidentally. Further, as the bank
keyboards retain the entered information until released, it is
possible to inspect the number prior to entry, and where a series
of numbers are to be entered in which only one or two digits are
changing, it is possible to let the numbers stand and change only
the va ry i ng digits.
2) Function Switches
Function Switches represent a form of selection device in which
the operator indicates to the machine that he wishes to make a change
and have the initiated action taken or not taken by the machine.
Usually the function switch Is a two-position switch, although it may be
a rotary switch or a multiple depression switch In which the color of
a light changes with each depression. Function switches may be used
singly or they may be used In groups whereby the selection of one
function switch from one group modifies a selection of another function
from another group.
A systems designer's greatest problem in the use of function switches
is usually where to put them. Since each switch represents an idea
or "concept communication" to the computer there are usually not
enough "finger holes" available to the operator to express all of the
ideas that he wishes to communicate. One approach which has been
taken to this problem by designers of command and control consoles is
to produce a matrix of switches, each of which generates the unique
code. This matrix Is covered by an overlay which identifies the
function of each switch within the matrix (See Fig . 2-22) » The matrix
overlay Is, itself coded in a manner that the computer can sense which
overlay is being used, and therefore, by first sensing which overlay
is being used and then sensing which switch is being depressed can tell
the function to be performed. In this manner, a 10 x 20 matrix of
switches with 100 overlays could be used to provide unique identification
of 2,000 separate functions. The obvious problem in such a system is
Start
Select
Geographic
Levels >,
_— .
Select
Owner
—
Select
User
—
—
PI
Select
Political
Levels
P7
P13
P19
P25
»
Subject
Install-
ations
Select
Output
Information
Content
Select
Output
Media
Color
Chip
P2
P8
P14
P20
P26
Select
Strike
Data
Fixed
Facilities
Totals
Printer
CRT
P3
P9
P15
P21
P27
Select
C/F
Estimates
Equipment
Degraded
Summary
List
Detailed
List
P4
PIO
P16 •
P22
P28
Select
Maximum
Dosage
Supplies
Residual
Tabular
Geographic
P5
Pll
P17
P23
P29
Select
Reporting
Personnel
Fallout
Intensity
Graphic
End
P6
P12
P18
P24
P30
—
■
RDA Output
—
—
I
Figure 2-22 Typical Display Overlay
2-3k
thai it takes an excessive amount of time to sort out the correct overlay
and pos i t i on it.
Another approach to the problem is to allow the computer to generate
a series of labeled boxes or points on a display, and allow the
operator at any time to select any one of these with a light pen or
similar device. In this manner, it is possible for the computer
to keep the operator continually informed of what switches it is
capable of accepting information from. Further, if there are a large
number of "overlays" that the computer uses, it is possible to allow
the computer to display a number or description for each overlay
allowing the operator to select the one which he wants and then choose a
switch on an overlay.
F i gure 2-23 shows a typical series of operator steps in using function
keys .
3) Pos i t ion I nd icators
A wide variety of position indicators suitable for computer
input are currently available. They include light pens, panagraphs,
etc. Most can be used either on-line or off-line. They depend upon
digitizing a series of points of a geometric figure.
2-95
Modify
1
Press Key
1
Computer Presents
a Display on the
CRT
Take Next
Directed
Action
Yes
End of
Sequence
Enter Data
into Display
Signal End
or Entry
Error
Make
Correction
Figure 2-23 Typical Series of Operator Steps in Using Function Keys
2-96
2.4.4.1.3 Human Language Interface
The present state-of-the-art of human language interface is currently
In the data collection stage. The primary group of devices in this category
is character recognition equipment used for alphanumeric and symbolic input
from printed and handwritten media. There are currently devices available
which can read limited fonts of printed data. Further work is anticipated
in this section during the next three months.
2.4.4.2 Machine-Man Interface
2.4.4.2.1 Visual Interface
Two forms of recorded visual interface are currently used as computer
output equipment, printed and graphical. Although it is possible for printing
equipment to produce graphical output in the form of a series of dots, bars,
etc., and for graphical output equipment to produce printing to label the
graph, they are separate and well defined classes of equipment which are best
separated for detailed study.
2.4.4.2.2 Printed Output Devices
Two basic types of printed output devices are available for use under
machine control. They are: the line printer, which produces a line of print
at a time, and the character printer or mechanized typewriter, which produces
one character at a time.
2.4.4.2.3 Line Printers
Line printers are computer output devices designed to provide a recorded
form of human language and symbolic language interface between machines and
man. They are designed to print one line of data at a time with the result
that a printing speed is dependent on the number of lines printed and independent
of the number of characters printed per line or of the total number of characters
printed. Such printers can be divided into four classes according to their
functional printing characteristics. These classes are:
2-97
1) Electromechanical
2) E lee tro-opt i ca 1
3) E lee t rograph i c
4) Magnetic
1) E lect romeciian i ca 1 Printers
Electromechanical printers are characterized by their ability to
produce carbon copies. The structure of these printers is such that the
paper is set between the controllable mechanical character forming device
(type) and a backing. These two are brought into contact at an appropriate
time creating pressure between them thereby transferring ink from a ribbon
or other source to the paper. This forms the character on the paper. Since
mechanical pressure is involved, this machine can produce carbon copies.
Since electromechanical printers depend upon an ink transf errence process,
it is necessary to somehow renew the ink supply. Moreover, unprepared papers
can be used with these printers.
To more readily explore the state-of-the-art of electromechanical
line printers, we may divide them into the following seven groups:
a) Rotating Drum Printers
b) Impact Wheel Printers
c) Matrix Printers
d) Sty 1 i St Pr i nters
e) Chain Printers
f) Stick and Rack Printers
g) Miscellaneous Printers
a) Rotating Drum Printers (Fig. 2-24)
The rotating drum printer is characterized by a solid drum or series of
wheels joined together on a shaft which contains one or more complete type
fonts for each column position to be printed. An inked ribbon is passed
slowly in front of the type font to provide the source of ink to be transferred
Paper is fed between this inked ribbon, and a hammer or actuator strikes the
back of the paper when the desired character is opposite the hammer position.
The pressure of the hammer is thus transferred through the piece or pack of
paper to the carbon ribbon and thus to the surface of the character on the
drum.
2-S8
Selection of characters is accomplished by indexing the position of
the drum and Tiring the hammer at the appropriate time to print the
desired characLer. Rotating wheel printers are characterized by clean^
high-quality impressions of individual characters. However, there is a
tendency for smear of the horizontal parts of letters and numbers at high
speeds. Such printers are plagued by more or less serious problems of
horizontal alignment as a result of timing differences between hammers.
Tape Wheel
or Cylinder
Type Font
Side View
Paper Movement
Paper
Carbon
Paper
Hammer and Actuator per
Printing Position
Ink Ribbon
Figure 2-24 Rotating Drum Printer
2-99
b) Impdct Wheel Printers (Fig. 2-25)
Impdct wheel printerb d re u clcjss of line printer commonly used in
addiny machines. Such printers are usually limited to numerics and a
few symbols, and they operate at relatively low speeds. in this class of
printer, a separate VN/heel is provided for each column position containing
all of the digits to be printed in that position. An indexed stop is used
to cause the wheel to stop rotating at a point so that the character to be
printed will be opposite the print position. All wheels are rotated until
they reach a stop position, at which time they are thrown forward against
a platen. Interposed between the type and the platen is a carbon ribbon
and the paper to be printed.
Tape Wheel
Paper Movement
Print
Motion
Platen
Side View
Ink Ribbon
Figure 2-25 Impact Wheel Printer
2-100
c) Matrix Printers (Fig. 2-26)
The matrix printer is a mechanism for impressing a number of "dots"
on paper to form a character. The dots are formed by the ends of wires
which are moved forward by energy supplied from an actuutor. These v/i res are
usually placed in a rectangular array causing the printing of a 5 x 7 dot
matrix (the smallest matrix which will print all alphabetics and numerics).
A character generation device must be used to determine the dots necessary
to print the selected character. The number of actuators required for this
approach is very large since each wire requires a separate actuator.
As the wires forming the character are fired against the paper through
an inked ribbon, the printing occurs from the front rather than the back as
with the wheel or cylinder printer. The result Is that such printers are
capable of producing a greater number of carbon copies than are printers
which require that the Impact be presented from the back of the pack of
paper. Ten or so carbons are usually considered maximum even with relatively
thin paper. The use of a sma 1 1 number of wires or dots to form a character
results in a low print quality; however, this may sometimes be partially
compensated for by the improvement in alignment that results from the
simultaneous firing of all wires. Since the character forming matrix is
external to the machine, a large number of actuators is required, and since
the wires that transfer the force to the paper are small and delicate, these
systems require a very high level of maintenance to stay In operation, and
they are very complex in their construction.
1 Character
Wire Matrix
5x7
Paper Movement
-' Paper
Carbon
Paper
Carriage
Side View
Ink Ribbon
Figure 2-26 Matrix Printer
2-101
d) Stylus Printers (Fig. 2-27)
The stylus printer, though an outgrowth of the matrix printer, is quite
different in its concept and performance characteristics. As v/ith the
matrix printer, a web of paper is passed over a carriage behind an inked
ribbon. Printing is by moving a series of styli horizontally between the
inked ribbon and a series of actuators (usually one actuator is used for
each character position). As the styli move horizontally across the paper,
the actuators press them against the inked ribbon at those points vvhere the
black part of a letter is crossed. The result is a line of characters composed
of a series of horizontal lines that are spaced closely together. The effect
achieved is much the same as that obtained by a television raster.
1 Character
Wire Stylus
Paper Movement
Paper
Carbon
Paper
Carriage
Ink EUbbon
Side View
Figure 2-27 Stylus Printer
2-102
e) Belt Jnd Chain Printers (Fig. 2-28)
Belt and chain printers are much like wheel and drum printers in
their configuration in that an inked ribbon is interposed between the type
and the face of the paper, and the character to be printed is selected by
firing a hammer against the back of the paper when the selected character
reaches that hammer position. The major difference between the two classes
of printers is that in the chain printer, type travels parallel to the
line of print, and in the wheel printers, type rotates perpendicular to
the direction of paper travel. Belt and chain printers are able to produce
about the same quality of print, the same number of carbons and with the
same speeds as wheel or drum printers. The horizontal movement of the type
reduces the horizontal alignment problem that results from the vertical type
movement of wheel printers; however, substituted for this is the problem
of vertical alignment. Specifically, the horizontal movement of the chain
tends to drag the paper in a direction of chain movement and thus pulls the
printing out of registration with the background printed on the paper.
Flexing required by the chain or belt limits the top speed that can
effectively be reached with the chain printer to somewhat below that which can
be reached by the wheel or drum printer in which the type does not flex.
Since the chain is travelling in a direction horizontal to that of the paper,
it must be at least twice as long as the total line length of the paper.
If the line length of the paper is 13 inches (130 characters) it follows
that the chain must contain more than twice this number of characters to
double back upon itself. The result is that the type font is usually
repeated several times on the chain.
Top View
Paper Movement Up
□
Type Font
Ink Ribbon
Paper
Carbon
Paper
1 Hammer and
Actuator per
Printing Position
Figure 2-28 Chain Printer
2-103
f) Rack and Stick Pri nters (Fig. 2-29)
Rack and stick printers are an early class of line printer widely used
in tabulating equipment and adding machines. Such printers use a bar of
type For each columnular position. This bar holds individual sprint loaded
pieces of type for each character to be printed in that columnular position.
During each print cycle, the bar of type is raised vertically until It
reaches a stop which holds It at the position of the character to be printed.
When all type bars have been raised to their print position, the print
hammers (one for each type bar) are fired against the type bars, thus
extending simultaneously selected pieces of type from the type bars. This
type impacts against a ribbon transferring ink to the paper which is supported
on a platen or roller. During the print cycle, no horizontal or vertical
movement of the paper or the type takes place. As a result, it is possible
to obtain accurate control of horizontal and vertical alignment upon the form.
Type impact is through the ribbon to the front of the form allowing a greater
number of copies than can be obtained with the back-hitting technique used
by the rotating wheel and chain printers.
Rack and stick printers are inherently slow as the stick or rack of
type represents a large reciprocating mass. This, combined with the large
number of moving parts, tends to require a relatively higher amount of
intenance per million lines printed than more modern types of printers.
ma
2-103 a
HAMMER
ZONING STEPS
SETUP PAWL
STOP PAWL
INSIDE BAR
OUTSIDE CASING
Type bar has been zoned
and raised. The Hammer
is firing against the "G"
position.
Figure 2-29
Stick-Type Printer Bars
2-104
g) Miscellaneous Printers
The h ypocyc loidi c printer, like the stick printer, forces type against the
face of the paper with no relative horizontal or vertical movement between
the type and the paper during the movement of impact. The type is contained
on a type drum in much the same manner as it i s on a wheel or drum printer.
Unlike the drum printer, the type cylinder of the hypocyc bid i c printer does not
revolve about its center line; rather, it is geared to provide surface motion
that advances a line of type perpendicular to the center of rotation, then
retracts and rotates one character position.
Although the drum is in continuous rotation at the moment of peak
advance, there is no component of relative rotational movement. The result
is that printing obtained from such a system exhibits no smear and, except
for paper wrinkling, will always be in excellent horizontal and vertical
a 1 i gnment.
Printing may be accomplished either by firing a hammer against the
back of the paper or by fixing a stop in position at some time prior to
the advance of the type. This stop may be fixed in a forward print position
or removed to a no print position. When the line of type advances to the
stop, it impinges upon the paper and presses it against the stop when it is
in a print position. The stop is not reached and thus no pressure is applied,
in a non print position. Such an arrangement requires some flexing in the
paper and does not lend itself to the use of a carbon ribbon. Instead, the
surface of the type is inked as it would be in a letter press. Conventional
inks dry and cake on the face requiring frequent cleaning, so aniline dye
i s usua 1 ly used ,
Hypocycloid ic printers are not well suited to printing many columns or
large type fonts since the requirement for strength in the central drive
shaft becomes too great. They can effectively produce a limited number of
columns of numerics or mixed numerics and symbols. Speed of such devices
is relatively slow because the internal drive shaft must make one complete
revolution for each character printed. Thus, to print at 100 lines per minute
from a 16 character type font, the central shaft must revolve at 1600
revolutions per minute, Hypocyclol d I c printers have found some application
In military situations due to their small number of moving parts and relatively
good reliability at low speeds.
2-105
In summary, the characteristics of electromechanical line printers
5 6 7
are compared in Table 2-2. ' '
2) Electro-Optical Printers (Fig. 2-30)
The electro-optical printers print by the projection of a direct optical
output onto a sensitized surface which is then developed to provide a printed
output. As the optical output and character generation equipment used is the
same as that used in displays, it will not be discussed in this section.
Probably the best example of an electro-optical printer is the
General Dynamics/Electronics SC7330 Printer. This printer is rated at
3000 to 5000 words per minute but can operate over a range of 10 characters
per second (100 words per minute) to 71,000 characters per second (on a line
basis). The particular printer is designed to print 128 characters per line.
The image generated on the face of the Charactron tube is projected
through an optical system onto a sheet of plasticized paper which has
previously been given a surface charge. Since the Charactron tube presents
the characters in serial fashion across the face of the tube, the character
presentation is asynchronous. The light generated from the phosphor is
projected through the optical system and falls on the charged surface of the
plasticized paper. This charge is held on the paper until it is advanced
through a "dusti ng"bath. At this point, the surface charge in the location
of the characters attracts fine particles of black polyethylene dust which
temporarily adhere to the surface of the paper. The paper is then advanced
at a fixed rate over a heating element that fuses the black polyethylene
to the surface of the paper, thus completing the printing process.
The process involved is very similar to the Xerographic process except
that the paper is directly charged. The light impinging upon its surface
can be used to "fix the charge" and thus attract the "ink" directly to the
area to be printed. In the Xerographic process, a selenium drum is used and
an electrostatic charge is placed upon it attracting the "ink" to the surface
of the drum. The image must then be transferred to an offset roller and then
to the paper itself where it is fused in place. Because of the offset nature
TABLE 2-2
CHARACTERISTICS OF ELECTROMECHANICAL LINE PRINTERS
Rotating
Wheel
Chain
Ma t r i X
Stylus
Stick and
Rack
Hypo-
Cycloidic
Impact
Wheel
Print Quail ty
Good
Good
Poor
Fa i r to
Good
Good
Good
Good
Vertical Alignment
Good
Fai r
Good
Good
Excel lent
Excel lent
Good
Horizontal Alignment
Fai r
Fai r
Good
Fai r
Good
Excel lent
Fai r
Number of Copies Produced
6
6
10
10
6
2
8
Speed-Li nes/Min
2000
1100
1000
300
150
300
150
(with indicated type font)
Type Font - No. of Characters
64
48
48
64
37
12
12
Type Font - Variable with
Yes
Yes
Yes
Yes
No
No
No
Change in Speed
Electrical Complexity
Low
Low
High
High
Low
Low
Low
Mechanical
Medium
Medium
High
Low
Medium
Low
Low
Maintenance Requirements
Low
Low
Very High
Medi urn
Medium
Low
Low
Advantages
High Speed
o
ON
2-107
of Xerography, il is possible to use any type of paper, Hov;ever, in the
process used in the General Dynamics/Electronics printer, a specially
p las t i c i zed-su rface paper must be used. V/e are informed, however, that
this paper is relatively inexpensive and has an indefinite shelf life.
Advantages of a printer such as this are;
1) No moving parts except the paper advance mechanism
2) Very high speed printing
3) Large type font possible (perhaps 200 characters) without
decrease in printing speed
4) Type font readily changeable by changing charactron tube
5) Very quiet printing
6) Long life, high reliability with low maintenance
7) Cost
8) Essentially asynchronous operation
9) Can be used to present graphical output
10) Printed output may be used directly as a multilith master
Disadvantages of Charactron Printer are:
1) Produces only original - do copies available
2) In its present form, machine is relatively heavy and bulky.
A similar printer to the General Dynamics unit is the Rank Printer
developed by Rank Precision Industries of England. This unit uses the
Xerographic principle and a standard cathode ray tube with a resistive
voltage divider in the deflection circuits to form individual letters.
2-108
Cathode Ray Tube
Paper
Remove Static
Charge
Fusing Unit
Optical System
Black
Developer
Powder
Figure 2-30 Electro-Optical Printer
2-109
3) E lectrographi c Printers (Fig. 2-31)
As yet, there has been little commercial exploitation of the electro-
graphic printer as a computer output device. This is probably due to the
requirement for special paper and the difficulties in producing multiple
cop ies .
The electrograph ic printer requires the use of specially coated paper
with high dielectric properties. This paper is moved across the matrix
consisting of wires imbedded In plastic. As the paper moves in front of the
matrix, It Is charged by the selected application of high voltage to the
matrix wires. The charged image is developed by running the paper through
a hopper containing a "toner" or powdered Ink in combination with dies
and thermosetting material. The "toner" adheres to the charged areas of
the paper and is then carried across the surface of the heating element
which fixes the Image by melting the thermosetting material enough to fuse
It to the surface of the paper.
Systems of this type have been built by Burroughs and A B Dick, The
Burroughs System employs a matrix of wires Imbedded in plastic In standard
5x7 form as the character generation media. The system Is able to print
at very high speeds, about one or two microseconds per character. As
recording takes place in parallel, paper feed becomes the major speed
1 imi tat I on.
The A B Dick electrograph i c printer uses a special matrix tube built
by the Stanford Research Institute. The tube consists of a cathode ray gun
aimed toward an assembly of fine wires imbedded in the glass face plate.
The electron beam is controlled by character-forming circuits external
to the tube. The wires provide a path for the charge from the beam to
flow outward to a special coated paper in front of the tube leaving the
character as an electrostatic charge on the paper. The use of the vacuum
tube Is considered a disadvantage for some applications; however, a much
higher resolution is obtained than can be obtained with the Burroughs System,
Thermal Treatment to Make
Printing Permanent
Heater
Contains Toner and
Carrier
Toner
Application
Electrostatic Printing
Tube
Wire Matrix
2-1 10
Paper
Paper
(Dielectric Coating)
Figure 2-31 El ect rograph ic Printer
2-
A third family of printers uses a Hogan facsimile stylus print head
and amplifiers. Printing is in the form of a 7 x I] matrix on a 10 x 15
dot field with 100 dots per inch.
It is possible for the electrograph i c printer to produce at least two
copies of the same document in a single printing, since the voltage applied
is high enough to pass through two sheets of paper simultaneously. As each
copy produced requires its own ink hopper and heating element, it is not
possible to quickly change from one number of ODpies to another.
Advantages of electrographic line printers are:
1) very high speeds
2) possibility of more than one copy
3) ability to form large type font
4) simplicity of electrical design
2-112
4) Magnetic Printers (Fig. 2-32)
Currently, there are no magnetic printers in production. However,
developmental work has been done by the General Electric Corporation,
Schenectedy; Univac Division, Sperry Rand Corporation, Philadelphia;
and National Cash Register.
The magnetic printer produces a shaped magnetic field vyhich is recorded
on a magnetizable surface. This surface is then exposed to some form of
magnetic particles which will be attracted to it and form the shape of the
magnetized character. This "inked" surface is brought into contact with
the paper and the ink is transferred from the magnetized surface to the paper.
The ink is fused to the paper and the magnetizable surface is then erased
for reuse.
The shaped magnetic field may be created by the use of a magnetized
type wheel, matrix or stylus, and the quality of output will depend both
upon the character-forming system used and the resolution obtained in
magnetization of the magnetizable surface. Like all transfer printing devices,
the magnetic printer can produce only original copy. All magnetic printers
require the use of magnetic material in the ink.
2-1 13
Shaped Magnetic
Field
Permanently
Magnetizable
Surface
Charged
Particles
Specially Prepared
Paper
Figure 2-32 Magnetic Printer
2-1 1^4
2.4.4.2.4 Character Printers
Character printers are computer output devices designed to provide a
recorded form of human language and symbolic language interface between
the machine and man. They are designed to print one character at a time
horizontally across a piece of paper. Printing speed is directly propor-
tional to the number of characters and control actions that must be taken
by the printing device. The use of this type of device usually requires
a number of special control functions and corresponding special control
codes. Typical of these are: space, bacl< space, precedence, e.g. upper
and lower case.
The character printer is usually used as a communications device, as
a part of a document originating device, as an output on an operator's
console, or as a very low speed output device. Character printers are
electromechanical in nature and are, therefore, capable of producing
carbon copies. All operate at speeds between ten and twenty characters
per second and use alphanumeric type fonts. For purposes of detailed
examination, electromechanical character printers can be divided into
five classes:
1) Typewriter
2) Drum Printers
3) Ball Printers
4) Matrix Printers
5) Stick Printers
2-1 15
1) Typewr i ters
One of the early forms of character printers was the modified
electric typewriter in which the keys were operated under computer control.
This type of device is still frequently used and is often found without
the keyboard operating solely as a printer. Most such devices are capable
of presenting a font of 44 characters. If a precedence code is used, 88
characters are available; however, 25 of these are usually lower case
alphabetics leaving a net type font of 52 characters. Most typewriters
operate at a maximum speed of 10 to 12 characters per second. This speed
is reduced by the time required for the execution of special function codes,
e.g. carriage return, back space, carriage shift.
In operation, the typewriter selects one of many levers, each containing
a character, and throws it into engagement with a power source in a manner
such that the character on the lever is fired against an inked ribbon
bringing it into contact with the surface of paper to be printed. The
array of type containing levers, or type basket, cannot be moved, w. th
the result that a carriage containing a platen and paiper must be moved
back and forth in front of the print position of the type basket. Dis-
advantages of such a system are that the paper must continually be moved
both horizontally and vertically, thus putting unusual stress on perforations
As the type basket is a rather complex mechanical arrangement of levers,
springs, clutches and pulleys, there are many moving parts that may fail.
An advantage of this system is the front impact hammer which allows many
carbon copies to be produced.
2) Drum Printers (Fig. 2-33)
The drum printer is an electromechanical character printer in which
the type is contained in one or more rows around the circumference, or
partial circumference, of a cylinder. This type drum is backed by a carbon
ribbon, the paper to be printed, and a hammer. The character to be printed
is brought into position by lateral and rotary movements of the type drum.
2- JI6
'I'VI'HWIIHHI.
RIGHT RESET ARM
PIVO'
-Ei-T RESET .ARM
-At
MM,
AMMER
LINK
PRINT BAIL
PRIN'J' HAMMER
BAIL
TORSION SPRING
PRINT TRIP
ARM
PRINT RESET
ARM
POWER BAIL
Figure 2-33
Drum Printer
2-1 17
When Lhe proper character is in position, the drum movemerit is stopped
and held in place until the hammer is fired, thus forcino the paper ond
carbon ribbon forward against the type.
In most present-day drum printers, the paper is held in position by
a paper transport mechanism, and the type drum and hammer are moved
horizontally across the paper, thus requiring no horizontal paper movement
3) Ba 1 1 Printers
Ball printer operation is similar to that of the drum printer except
that the characters are formed on the surface of a ball. The character
to be printed is selected by a combination of vertical and horizontal
rotary movements of the ball bringing it into the selected print position.
When the character i s In position, the ball is driven against a carbon
ribbon printing the character. Character isolation obtained through the
use of a ball allows elimination of the back hammer reducing the inertia
of the system and allowing faster operating speeds. It allows a front
impacting system which produces a greater number of carbon copies. Like
the drum printer, the ball printer requires no horizontal movement of the
paper, thus contributing toward systems reliability. Ball printers are
capable of operating at rates up to 20 characters per second due to their
low Inertia. Like the drum printer, the ball printer readily lends
itself to changes In type font.
4) Matrix Printers
The matrix printer, like the ball printer and typewriter, is a
front printing device. Spring loaded type is held In a rectangular
matrix in front of a carbon ribbon. Characters are selected by horizontal
and vertical movement of the matrix which brings the selected character in
line with a hammer. When in proper position, the hammer is fired against
the character, thus giving It the Inertia to Imprint on the paper. Like
the ball printer and the drum printer, the matrix is carried horizontally
across the face of the paper requiring no horizontal movement of the paper
i tself .
2-1 18
The most common application of the matrix printer is in teletype
operation where it is operated in the range of five to seven characters
per second.
5) Stick Printers
The stick printer is similar in concept to the drum printer. The
type is held on the face of a hexagonal or octagonal type stick. This
type stick is moved horizontally in front of the hammer and rotated to
bring the desired character into print position. When in print position,
a hammer fires against the rear of the paper bringing it into contact with
the carbon ribbon and the selected character. Usually, this type of printer
relies more on linear motion than rotary motion in the selection of the
character.
In summary, all present electromechanical character printers depend upon
selecting a character and bringing it into a fixed position in front of a
carbon ribbon and paper array; then applying the necessary force to transfer
an image to the paper. In all cases, the character is not in motion at the
time of transfer. The result is that a clear image can be obtained from
this type of printing. However, this is done at the expense of possible
improvements in speed.
In summary, the characteristics of electromechanical character printers
are compared in Table 2-3.
TABLE 2-3
ELECTROMECHANICAL CHARACTER PRINTERS
Typewriter Drum
Ball
Type
Ma t r i X
Type
Stick
Pressure Source
Character
Front
Hammer
Back
Character
Front
Character
Front
Hammer
Back
Changeable Type Font
No
Yes
Yes
No
No
Mechanical Conplexity
High
Med i um
Low
Med i um
Med i um
Speed Char/Sec
12
10
20
VD
2-120
2.4.4.3 Machi ne- to-Machi ne Interface
In current technology there are two broad classes of machines which
must communicate with each other and with others of their own type. The
first of these classes is the analog machine which Is an operating simile
or analog of the problem being studied. In an electrical analog machine
the data is portrayed by a voltage level. The voltage level is continuously
varied to correspond to the variations that occur in the function being
represented In the real world. The digital machine works with a mathematical
model of the real world and expresses all changes as changes to the magnitude
of a number In an equation.
To provide communication from an analog machine to another analog machine
It is necessary to convert the voltage level of the first machine into a voltage
level in the second machine which would represent the same number.
To allow one digital machine to speak with another digital machine, it
is necessary for the first machine to convert its digital representation
into the code representing the same digit for the other machine.
To allow communication between an analog and digital machine It is nec-
essary for the voltage level to be digitized, or for the digital representa-
tion to be converted into a voltage that can be recognized by the other
mach i ne.
Provided that two digital machines or two analog machines are designed
to operate at the same degree of accuracy, it is possible for them to
communicate without loss in the accuracy of the data as no rounding effect
Is Involved. Whenever a digital machine must speak with an analog machine,
or vice-versa, a conversion problem is Involved and there is some potential
o
loss of accuracy of the data.
2.4.4.3.1 Analog Machines
The problem of analog-to-analog communication has not yet been Investi-
gated and will be Included in this section at a later date.
2-121
2.4.4.3.2 Ana log- to-Digi ta 1 and Di g i ta 1- to-Ana log Machines
This area is currently under investigation; however, current state-of-
the-art is not fully explored.
2.4.4.4 Machi ne- to-Mach ine Communi cation--Diqital
There are two classes of digital mach i ne- to-mach i ne communication.
Real-time transmission and stored data transmission. In real-time trans-
mission concept, data is originated by one machine and transmitted to a
second machine which has responsibility for monitoring the line and
accepting data as it occurs. (See Figure 2-34). The transmission may be
bit serial, character serial, or word serial. Some form of buffering is
usually required either on the part of the initiating machine or on the
part of the receiving machine. This may take the form of a separate buffer
or a buffer that is an integral part of the process. Many problems are
involved in scanning a number of such machines by one machine, that is,
accepting the data in an order that will assure that no data is lost.
This section will deal with a wide variety of mach i ne- to-mach i ne communication
including console- to-computer communication, computer-to-console communication,
computer- to- computer communication, radar- to- computer communication, weapon-
to-computer communication and computer-to-weapon communication, etc.
2.4.4.4.1 Digital Stored Data Communication
In the concept of digital stored data communication, one machine
communicates with another machine which is a storage device or "data sink".
At some later point in time the same machine, or another machine, reads
data from the storage device. The concept is very much like that of using
an auxiliary memory, except that communication implies the ability in some
manner to remove the data from the machine and place it on another data
reading machine. It also implies that the data be read by the second
machine in the same manner in which it was recorded by the first machine.
2-122
Terminal
Terminal
Terminal
Terminal
Unit
Unit
Unit
Unit
>
k
/
k
> ^
i k
' <
> '
< '
Scanner and Multiplexor
1
>
k
>
'
\
'
1 Assemble
Storage
Assemble
Storage
1
I
/
^
Computer Input and Output
^
t \
r
Computer
Terminal
Multiplexing
Storag^e and
Buffering
Figure 2-34 Schematic of Typical Interface
Functions in Digital Machine-
to-Machine Communication
2-123
This functional difference separates data storage communication devices
such as magnetic tape, paper tape, and punched cards, from the random
access storage devices such as core memories, disc files, and drums. This
separation is a very real one in practice even though it is possible to
build devices with characteristics that are suitable in both applications,
2.4.4.4.2 Punched Cards
Where a unit record storage or communication is required, punch card
equipment represents the most economical and most efficient form of exist! ig
storage. Card punching equipment is available that will record at rates from
one to 300 characters per second, and card reading equipment is available
that will read from rates of 1 to 1000 characters per second. Although the
cost per bit of storage in this media is relatively high, it can well be
justified in certain types of application such as programming and document
handling. As the punched card represents a unit record, it is a particular
convenience where one machine must create data which must pass through
human hands and later be entered into a machine system. A current contender
for this field is a combination of printers and character recognition equipment,
However, in most applications the punched card is more economical on a per
9
bit handled basis. Detailed examination of present state-of-the-art and
developments in punched card equipment is currently underway and will be
included at a later date.
2.4.4.4.3 Punched Paper Tape and Incremental Magnetic Tape
Punched paper tape and incremental magnetic tape represent the
current state-of-the-art in incremental recorded continuous records.
Neither is particularly suitable to the unit record concept. Neither is
well suited for document recording. Both are capable of accepting or trans-
mitting character oriented data in an asynchronous fashion without regard
to block length. As a result, these devices are useful for interfacing with
communications type equipment. They can accept data from a relatively high
speed source and record and transmit asynchronously or continuously at a
lower speed. They can accept it from an asynchronous or low speed device.
2-124
and store it for high speed reading by a faster device. Both record data
in a continuous form and in a compact manner such that the data stores itself,
e.g. the reel of tape can readily be transmitted from point-to-point without
loss of data and without possible change in the ordering of data. The
equipment involved to record or read the information is relatively inexpensive
and easy to integrate in a system. Currently, a wide range of paper tape
equipment is available with recording speeds varying from one to 150
characters per second, and reading speeds varying from one to several
thousand characters per second. Although incremental magnetic tape equip-
ment is currently available, it is quite new and will, therefore, be
discussed in more detail under available input-output devices in the 1970-
1980 period.
Detailed examination of present state-of-the-art and developments in
paper tape and incremental magnetic tape is currently underway and will be
included at a later date.
2.4.4.4.4 Magnetic Tape Recorders
The magnetic tape recorder as used on a modern computer, provides both
a temporary storage device for the computer itself and a medium of trans-
mission of data from one point to another as does paper tape. It can be
used at a variety of speeds and can be used as a speed translation device.
In its conventional form, it is unlike paper tape equipment in that it is
inherently block oriented.
Although it is possible to have a block one character in length and
thus turn the tape machine into a character serial device, this is not a
practical piece of equipment. Equipment performance is based primarily
on start time, (that is, the length of time it is necessary for the tape to
achieve the necessary read/write speed) and a combination of the speed at
which the tape is passed and the bit density recorded on the tape.
Detailed examination of present state-of-the-art and developments
in magnetic tape equipment is currently underway and will be included at
a later date.
2-125
2.4.5 Availability of Input Devices in 1970-1980
As the input-output study is still in the evaluation stage, any
conclusions drawn in this section must be considered as preliminary.
Further evaluation will produce much more meaningful content for this
section and continuing investigation may disclose new and hitherto unknown
concepts which could result in usable input-output devices in the 1970-
1980 period.
2.4.5.1 Man-machine interface
Since little can be done to expand the information output rate of the
human, improvements in a man-machine interface must necessarily come from a
better utilization of the data passed through this channel. This, in turn, must
come about through a more sophisticated man-machine relationship In which the
data flow from the man to the machine expresses concepts which are common to the
man by virtue of his learning and are common to the machine by virtue of Its
program. In such a manner. It Is possible for the man and machine to establish
a "rapport" or working relationship in which, under certain circumstances, only
a small amount of information need be passed between the two: the machine
assuming what the man will do under certain circumstances, and the man assum-
ing what the machine will do under certain circumstances, thus relieving the
load on the Interface.
2.4.5.1.1 Sound
Although there Is currently much investigation under way as to means of
meaningful interpreting the sound patterns produced by humans; and, in spite
of the fact that there have been built a number of machines which have been
capable of analyzing human originated sounds and distinguishing those sounds
from a small vocabulary of words or phrases. It seems unlikely that practical
working voice Input equipment will be available for use in a tactical data
system of the 1970-1980 era.
2-126
2.4.5.1.2 Pressure or Motion
Human or language oriented keyboard devices are currently quite highly
developed and it seems unlikely that any new breakthroughs will occur in this
area. Such devices have, for many years, allowed the human to express himself
in his own native languages in a speed in excess of his capability to do so,
2.4.5.1.2.1 Graphical Input
Within the last few years, much work has been done in the development
of graphical input devices, and there now exists a number of such devices in
laboratory form. From the human standpoint these devices have proved to be
a rapid and convenient method of transferring graphic concepts from the man
to the computer. Unfortunately, as the computer is much better at algebra
than geometry, all such concepts must be handled mathematically and a great
deal of software must be prepared before the data transmitted through
such a graphical device can be meaningful to the computer. Much work has
been done in this area, some of it quite successfully. However, the
development of specialized software of this type has proved to be a complex
and painful process. it is anticipated that such graphical input devices
and their associated software will be available for use in a 1970-1980
system and will be a powerful adjunct to the conventional man-machine
i nterface,
2.4.5.1.3 Human Language Interface
The ability of input equipment to sense a human language representation
has been progressing slowly but steadily over the last ten years. Currently,
such equipment is available, and without doubt more sophisticated forms will
be available in the 1970-1980 period. The problem faced is the degree of
sophistication available in this type of equipment. To date, most equipment has
been built to order, with the result that there has been little requirement
for versatility, and hence the cost of the equipment has been very high. If
the development of a mass market occurs within the next few years, we can
foresee a substantial increase in the versatility of human language interface
equipment as well as a substantial decrease in its cost.
2-127
2.4.5.2 Machine-Man Interface
Like the man-machine interface, the machine-man interface is limited
largely by the ability of the human to accept a given data rate with a
given language. Since the modes of data reception of the human are quite
limited and well defined, the only hope for improvement in the machine-man
interface is In the improvement of languages of communication between the
machine and the man. The last ten years has seen a tremendous increase in
the variety of machine-to-man languages. At one time it was necessary for
the man to learn the machine language and think and speak in it. Today it
is possible for the machine to use some reasonably human oriented languages
for human output. It is quite likely that the next five to ten years will
see an increase in language availability of considerably greater significance
than that during the last five or ten years. If this occurs we can expect
the man-machine interface to become as efficient as the present man-to-man
interface. It is, in fact, likely that because of the machine's predictabil-
ity it is possible for the machine-to-man interface to be better than that
of the man-to-man interface.
2.4.5.2.1 Man-Machine Interface Visual
Current information indicates no major breakthroughs in the non-display
visual interface. There is currently a wide variety of equipment available
that will provide graphical presentation of spoken languages. Likewise, there
are a wide variety of plotters available to produce graphical output.
The speed range of the so-called high-speed printer used by the computer
industry may increase somewhat as a result of improved engineering. The type
font and speeds available on photocompos i ng equipment will probably improve
slightly, and there will probably be a substantial decrease in cost. Also,
plotters will probably become somewhat faster and somewhat more versatile,
2.4.5.2.2 Auditory
In the past few years there has been a rapid rise in the development
of equipment which is capable of selecting a pre-recorded audio message and
presenting it upon a digital command. There has also been some work on
equipment that is capable of selecting a variety of words and phrases and
2-128
combining them into a meaningful message. As yet, there has been little
such equipment put "on line" with a computer system. This equipment has
manifested itself as automatic airline call systems and selectable sales
message systems. There appears to be no reason whatsoever why such equip-
ment cannot be used as a machine-man interface under computer control. In
civilian life, there has been a good deal of reluctance to "take orders
from the machine". Use of the "open loop" concept in the military does not
include the possibility of allowing the machine to present decisions to a
commander and allowing him to select from those decisions which, if any, he
chooses to make. Having selected a course of action, there is no reason
why he cannot allow the computer to issue instructions in a vocal manner
just as he might allow it to make up instructions, e.g. orders, texts, etc.
in a wr i tten note.
2.4.5.3 Machine-Machine Interface
As with current equipment, two types of machine-machine interface
will be available. These will be the real-time message interface and the
stored message interface. The real-time message interface may occur through
any media that is capable of transmitting coded energy. In all likelihood,
electromagnetic data transmission will continue to be far more economical
than any other means except in special applications. While the columns per
disc may go down as the result of an increasing understanding for the nature
of data and an improvement in the design of equipment suitable for handling
data, it is unlikely that any new data transmitting technique will be found.
There are, however, excellent possibilities for the improvement of existing
techniques; largely through the study of the nature of data and improvement
In coding techniques to allow the transmission of a much greater amount of
data on a given bandwidth. Such forms of data compression are most promising
under some types of application. It must be kept in mind, however, that
they are of greatest usefulness where the total data received can be
integrated to achieve an over-all "effect" and where no single bit of data
is of vital importance.
2-129
As contrasted with data compression techniques, there will probably
also be a substantial improvement in the method of coding data to include
a greater amount of redundance and, therefore, allow a greater accuracy of
transmission and interpretation.
2.4.6 Limitations of Present and Planned Input-Qutput Technology
The review of this section is not complete at this stage. It will be
included in the final report.
2.4.7 Recommended Developments to Meet ANTACCS Requirements
No work has been done on this section as yet as the ANTACCS requirements
have not been made available.
2-4.8 Evaluation Criteria Recommended
Evaluation criteria are being established as a part of the evaluation
of technology which is currently in process. When requirements information
is available, the evaluation criteria for each of the separate technologies
will be presented in this section and related to the ANTACCS requirements.
2.4.9 Conclusions and Recommendations
As the input-output technology task is now only 50% complete, it is
possible to classify the technologies involved list sources of information
and provide limited information on current status and anticipated availability
in the 1970-1980 period. However, it is premature to attempt to evaluate
their limitations and suitability for the as yet incompletely stated ANTACCS
requirement. It is also too early to make recommendation as to future
technological developments to provide evaluation criteria or to form
recommendations or conclusi onss.
2-130
REFERENCES
1. A Method of Voice Communication with a Digital Computer, S. R. Patrick
and H. M. Willett, Proceedings Joint Computer Conference, New York, N.Y.,
Vol. 18, pp. 11-24, Dec. 13-15, 1960.
2. System Reads Three Type Fonts, J, M. Carroll, Electronics, pp 49,
Dec. 20, 1953, McGrawHill Publication.
3. Programming Pattern Recognition, G. P. Dinneed, Proceedings Western
Joint Computer Conference, Los Angeles, California, March 1-3, 1955
4. Digital Data Communication Techniques, J. M. Wier, Proceedings of the
IRE, Vol. 49, No. 1, pp. 196-209, Jan. 1961.
5. Printing Equipment for Medium, Intermediate, and Large Size Computers,
Staff of Cresap, McCormick and Paget, Control Engineering, Jan. 1962,
pp. 91-95.
6. Digital Printers, Editorial Survey, Instruments and Control Systems,
Vol. 32, May 1959, pp. 700-707.
7. Tape Printer Applications, W. R. Bea 1 1 , Instruments and Control Systems,
Vol. 32, May 1959, pp. 708-709.
8. Analog Input and Output System for a Real-Time Process Control Computer
System, C, A. Walton, Joint Automatic Control Conf erence--Proceed i ngs
13, 4. 1-4,6, June 1962.
9. Punched Card as Information Carrier and as Technical Problem, K. Lindner,
Feinwerktechnik, 67(2):55-61 1963.
10. Digital Circuit Techniques for Speech Analysis, G. L. Clapper, IEEE--
Transactions on Communications and Electronics 66:296-304, May 1963.
11. Minutes of ASA Committee X3.1 (Optical Character Recognition and its
Subcommittees, American Standards Association, Sectional Committee X3
on Computers and Data Processing.
2-131
2.5 MEMORIES
2.5.1 Classification of Memory Types
Memories can be classified in four functional categories based upon
their use and purpose in the system. Each of these categories requires
different combinations of speed, capacity, cost and other characteristics.
A specific memory technology may overlap two or more categories but the
relative importance of different characteristics and, to some extent, the
design approaches and criteria vary from one category to another. The four
categories used in classifying memories in this report are:
Registers and high-speed control memories
Main high-speed internal memories
On-line auxiliary storage
Off-line auxiliary storage
The high-speed control memory and the high-speed internal memory are
normally considered an integral part of the computer or central processor,
while the on-line auxiliary storage and the equipment for reading and writing
and the off-line auxiliary storage are frequently considered external
peripheral devices. However, the techniques used in mechanizing registers
and high-speed memory in the central processor are also used for control
and buffering functions in special purpose peripheral devices such as
communications terminals and certain types of input/output equipments.
This is illustrated by the use of the magnetic core matrix memory as
a small capacity high-speed control memory, as a large capacity high-speed
main internal memory, as a very large capacity slowrspeed on-line
auxiliary storage, or as a small capacity slow-speed buffer in a display
unit or a communication terminal equipment. Although a magnetic core
matrix memory would be used in each of the above applications, the
combination of characteristics designed into the core memory would be
quite different for each of these applications.
The auxiliary storage category represents very large capacity bulk
storage that is addressed by the computer in large blocks rather than by
individual words. It usually has an average access time several orders of
magnitude slower than that of the high-speed internal memory. The on-line
auxiliary storage is directly available to the computer under computer
control without manual intervention. The off-line auxiliary storage
2-132
normally requires a manual operation to place the storage media on the read/
write device (e.g. a magnetic tape unit) that is controllable by the computer.
In this sense, a magnetic tape unit can be considered an on-line auxiliary
storage if a particular reel of tape is written, left on the tape unit, and
later read back by the computer. However, if a tape reel is written by the
computer, taken off the tape unit, and stored on the shelf to be later put
back on a tape unit to be read into the computer again, it would be considered
off-line auxiliary storage.
Most off-line storage equipments such as magnetic tape units, punched
paper tape equipments, and punched card equipments are commonly classed as
input/output equipment since they act as input/output devices to the central
processor. Unfortunately, this tends to obscure the fact that these devices
are actually serving an off-line auxiliary storage function in the overall
system rather than an input/output function. They should not be confused
with "true input" and "true output" devices such as keyboards, printers,
ana log- to-d ig I ta 1 converters, and digital-to-analog converters. In this
report, these types of off-line storage devices are listed in their con-
ventional category as input/output equipments and have been discussed
previously in Section 2.3.
Equipments such as magnetic card memories and magnetic disc files with
removable disc stacks are on-line auxiliary storage devices with respect
to the cards or discs actually on the device at a given time. However,
they act as read/write and access mechanisms for off-line storage with respect
to the cartridges of magnetic cards or the stacks of removable discs on a
shelf if these have been written by the device previously and will be read by
the device subsequently. In this report, these devices are discussed in the
category of on-line auxiliary storage. Those that can act as read/write mechan-
isms for off-line storage will be identified.
In addition to four major categories discussed above, certain memories
can be classed in a fifth ca tegory--"specia1 memories." This category would
i nclude:
Read-only memories
Read-mostly memories
Associative memories
Analog memories
2-133
While some of these special memories may be mechanized with completely
different techniques than those in the other categories (e.g. photographic
read only memory), others may be mechanized with essentially the same basic
techniques but with additional control and logical functions or, in some cases,
with certain hardware omitted (e.g. thin-film read only memory).
2.5.2 Sources of Information
2.5.2.1 People and Organization
Memories have been discussed with personnel of a number of different
companies and governmental agencies in the course of this study. The following
list indicates the companies and governmental agencies with whom memories have
been discussed and the type of memories duscussed with each:
Rome Air Development Center
Rome, New York.
Control Data Corporation
St. Pau 1 , Mi nnesota
Remington Rand, Univac
St. Paul , Mi nnesota
Bunker'-Ramo Corporation
Canoga Park, California
Fabr i tek
Emery, Wisconsin
NCR
Hawthorne, California
Sylvan ia
Waltham, Massachusetts
Philco Research Lab.
Newport Beach, California
Cryogenic Memories,
Random Access Memories,
Non-Rotating Mass Memories,
Cryogenic Associative Memories,
Magnetic Associative Memories
Magnetic Thin Film Memories
Coincident Current Magnetic Core Matrix
Mems., Word Organized Magnetic Core
Matrix Memories, Magnetic Thin Film
Memories, Associative Memories
Woven Screen Memories,
Biax Associative Memories,
Thin Film Plated Wire Assoc. Memories,
Transfluxor Associative Memories,
Coincident-Current Magnetic Core Matrix
Memor ies ,
Magnetic Thin-Film Memories
Plated Magnetic Rod Memories
Photochromic Memories
Permanent "uni t- record" magnetic
card memories
Magnetic Thin-Film Memories,
Biax read-mostly memory,
Biax associative memory
2-134
RCA Laboratories
Princeton, New Jersey
Laboratory for Electronics
Boston, Massachusetts
Autonetics ,
Anaheim, California
U.S. Army Engineering
Research and Development Lab.,
Fort Monmouth, N.J.
Office of Naval Research
Washington, D.C.
National Security Agency
Fort Mead, Maryland
U.S. Navy Bureau of Ships
Washington, D.C.
Hughes Ground Systems Division
Fullerton, California
Cryogenic Random- Access Memories,
Cryogenic Associative Memories,
Laminated Ferrite Memories
Permalloy Sheet Random- Access
Memories ,
Bernoulli-Disc Rotating Memories,
Bernoulli-Disc Mass Memories
Memory Systems and Hierarchies
Block Oriented Random- Access Memories,
(static ferro-acoust i c storage),
Memory Systems and Hierarchies,
Magneto-optical and Electro-optical
Memories for Displays,
Glass delay-line Memories
Magnetic Core Memories,
Read only Memories,
Cryogenic Memories,
Associative Memories,
Optical Memories
Thin-Film Memories
Cryogenic Memories
Associative Memories,
Magnetic Dl sc Fi les,
Magnetic Card Memories
Associative Memories,
Magnetic Domain Wall Shift Registers
Much of the information presented in subsequent parts of this section is
based on discussions with memory experts in the organizations listed above.
In addition to discussing memory techniques that have not been adequately
described in published literature, the opinions of these experts concerning
the advantages, disadvantages, limitations, and future of different memory
techniques were solicited.
2-135
2.5.2.2 Literature
An extensive list of references pertinent to the study of memory
technology are given in the Bibliography. Most of these references have
been scanned, but only a limited number of the more important ones have been
studied in detail to date. Some of the material presented in subsequent parts
of this section on memory technology has been extracted from some of these
references. In some cases, where noted, direct quotations have been used.
During the remainder of this study, the more pertinent and important of these
references will be studied in detail and new references reflecting materials
published or discovered subsequent to the preparation of this Bibliography
will be i ncluded.
2.5.3 Memory Characteristics for ANTACCS
The characteristics required for memories in ANTACCS cannot be fully
identified at this time since the results of the requirements analysis have
not been available. However, it is certain that registers and high-speed
control memories, main high-speed internal memories, on-line auxiliary
storage, and off-line auxiliary storage will all be required in a future
Naval Tactical Data System.
It is anticipated that the results of the requirements analysis and further
studies in the area of machine organization will indicate whether requirements
exist for any special memories.
In the way that the study of memory technology is being handled, it
will not be materially affected by the results of the requirements analysis
except in some detailed areas such as environmental conditions. All of the
more important and more feasible types of memories that might be applicable
to 1970 systems of either the ANTACCS or AMTACCS types are being analyzed
and will be evaluated and compared. Memories meeting only commercial
specifications and those for aerospace requirements are not being considered,
but the memory technologies being studied will cover the full range of
requirements for shipboard and ground-based military environments. Types of
memories required for particular functions in a computer, or in other parts
of the system, can be selected based on the characteristics that will be listed
2-136
and compared. The memories to be stud led, compared , and evaluated will
encompass all types necessary to meet all of the functional requirements
in these sys tems--both in the central processor and in peripheral and
auxiliary equipment.
2.5.4 Applications of Memories in the Naval Environment
It is anticipated that all memory requirements for shipboard and ground
based military environments will be covered by this study. Applications of
memories in ANTACCS and AMTACCS systems will include high-speed control
memories, high-speed internal memories for program and data storage,
on-line and off-line auxiliary memories for storage of multiple alternate
programs, large data files, and historical records, control registers and
buffers for input/output equipment, communication terminal equ ipments, and
display equipments and status registers. The applicability of specific
types of memories to different applications will be considered in further
detail in the remainder of this study.
2.5.5 Current Status Review
2.5.5.1 Specific Memory Techniques Investigated to Date
In addition to the conventional magnetic core matrix memories, magnetic
surface storage memories (e.g. tapes, drums, discs, cards) and delay line
memories, the following types of memories have been investigated so far
during the study:
Continuous-sheet cryogenic random access memories
Cryogenic associative memories
Magnetic associative memories
Planar magnetic thin- film memories
Cylindrical magnetic thin-film memories
Woven screen memories
Permanent magnetic unit record card
The laminated ferrite memory
The flute memory
The permalloy sheet memory, and
Ferro acoustic memories
A brief description of each of these memory techniques is given, and the
information collected to date Is summarized in this section. However, these
memories have not been analyzed, compared, and evaluated in detail at this
time.
2-137
2.5.5.2 Continuous-Sheet Cryogenic Random- Access Memories
Cryogenic memories are based on the principle that at temperatures
near absolute zero degrees, the resistance of certain materials (super-
conductors) may be either zero or some finite non-zero resistance depending
upon the magnitude of the magnetic field surrounding the superconductor. In
one type of continuous-sheet cryogenic memory under development by RCA
Laboratories, the storage media is a superconducting tin film. This tin
film, a lead sense line, and lead drive strips are fabricated by vacuum
deposition techniques on a two- i nch square substrate. These metallic
films are all insulated from one another by vacuum deposited insulating
films (silicon monoxide). The sense line is beneath the storage plane and
is oriented diagonally to and directly under the intersections of the two
sets of drive strips which are orthogonal to one another.
Coincidence of the "X" and "Y" drive currents at an intersection of
the drive strips produces a magnetic field sufficiently strong to switch the
region of the storage plane beneath the intersection out of the superconducting
state, thus permitting magnetic flux to link through the plane in that region.
When the currents are removed, the flux linking the small region of the
continuous sheet is trapped, and persistent currents are established in the
storage plane to support this trapped flux. The stored information can be read
by subsequently applying a coincidence of drive currents of proper polarity, and
then sensing the voltage change on the sense line.
In a later refinement, a cavity sensing technique is used rather than
a zigzag sense line. A second continuous tin film is located a slight distance
beneath the storage plane and connected to it electrically along one edge.
In this case, when the proper polarity drive currents create a magnetic field
sufficient to destroy the superconducting state in the region of the inter-
section, a change of flux is created within the cavity beneath the intersection
between the storage plane and the sense plane. This causes a sense pu.lse
at the output tabs connected to the sense plane.
2-138
This memory is made by batch- fabri cat i on techniques. The continuous
sheet superconducting storage film and sense film and the X and Y drive strips
are vacuum deposited for an entire plane. In addition, the X and Y selection
matrix is mechanized by means of cryotrons that are vacuum deposited on the
same plane and connected to the drive strips by the deposition process.
A 16,384 bit memory plane, including 508 cryotrons for XY selection,
is fabricated in 26 vacuum deposition steps using 16 different masks. The
planar density is approximately 10,000 bits per square inch. The anticipated
characteristics for this memory are:
^ to 1 microsecond access time per word
10° bits in a 10-inch square plane
10^ bits in a complete memory
production capability during 1964
300 ma selection current
100 ma drive current
10^ bit memory will cost approximately $1,000,000
A 16,384 bit plane (128 x 128) with cryotron addressing on the same
2-inch square cryogenic plane has been in operation at this time.
RCA considers the next step to be a 256,000 bit plane with 0.0005 inch
spacing between wires on a four- inch square plane. Planes of 1024 x 1024 are
an eventual goal. By 1965, RCA anticipates that they will have a 10 bit
prototype memory.
Because of the basic overhead cost in the refrigerant, the cryogenic memory
is not considered a good approach to a small memory, but becomes more advan-
tageous as the size increases. RCA believes that the cryogenic approach is
the more likely way to reach capacities of lo" bits, but that the laminated
ferrite memory will be better for capacities of 10' bits. The crossover
7 R
between the two will occur somewhere between 10 and 10° bits. The present
cryogenic memory design is limited to cycle times of 1-2 microseconds because
of the cryotrons used In the selection matrix.
To date, RCA has fabricated two good arrays (128 x 128), but both of
these have been damaged by repeatedly taking them in and out of the liquid
helium. They are now working on ten arrays for the Air Force.
2-139
2.5-5,3 Cryogenic Associative Memories
Work on cryogenic associative memories has also been underway at several
2 3
companies, ' The cryogenic approach is probably worthwhile only for very
large capacity associative memories. The cryogenic associative memory under
development at RCA is essentially a two "core" per bit type using two parallel
continuous sheet planes. The associative feature depends upon whether a cor-
responding spot in both planes is storing the same information. If they are
not, the flux between the two planes cancels out. The RCA continuous sheet
cryogenic memory uses lead for the super-conductive wiring, tin for the
continuous-sheet superconductive storage planes and silicon monoxide for the
insulators. These are all deposited in approximately 20 different deposition
steps with different masks for each. Fabrication is currently mechanized,
with a circular jig holding the 20 masks in the vacuum so that each mask can
be rotated into place at the proper time.
2.5.5,4 Magnetic Associative Memory
The term associative memory is used to refer to a memory which is addressed
by content rather than by a unique numeric address. An associative memory
involves sufficient logical capability to permit all memory locations to be
searched essentially s imul taneous ly-- i . e, within some specified memory cycle
time. The search may be made on the basis of the entire contents of each
location or upon the basis of selected bit positions of each location. Thus,
it is possible to search for all words meeting certain criteria or for all
words in which a certain tag portion of the word meets the criteria. Searches
may be made on the basis of equality, greater- than-or-equal- to, less- than-or-
equal-to, and between limits.
Associative memories are also referred to as content addressed memories
and search memories. In a sense, all of these terms are misleading In that
the unit is not an associative "memory" but rather an associative "processor"
since the memory function involves a minor portion of the total hardware and
costs. The major portion of the hardware and costs is that involved in the
logical operations necessary to accomplish the parallel search of memory
locations. For example, a magnetic associative memory frequently requires
a sense amplifier for each word location, and logical capability for each
word (in some cases for each bit portion) to permit the parallel search
2-140
capability. These hardware elements are not required for the memory function
itself. The memory function alone might account for only 1/4 or 1/5 of the
total costs based on comparing the costs of the associative memory with that
of a random-access memory with equivalent capacity and access time.
An associative memory differs from most random access memories from a
hardware standpoint in the following ways:
1) The sense lines are word oriented
2) The drive lines are bit oriented
3) There Is a large number of sense amplifiers as a result of
the word orientation of the sense line
4) The sense amplifiers have a high duty cycle which tends to
make them more expensive
5) A large amount of logical circuitry is required
The costs of the sense amplifiers and logical circuitry are the main
factors contributing to excessive costs of associative memories compared to those
of random-access memories. An associative memory also requires a non-destructive-
read-out storage element since it is not feasible to rewrite every word location
for each search.
Although some magnetic associative memories use a single aperture magnetic
core, it is convenient to use a multi-aperture device such as the Transfluxor
or the Biax element. The Transfluxor approach requires two elements per bit
and requires a greater drive signal since it is necessary to actually change
the direction of saturation of the material around the interrogate hole. The
Biax element requires a relatively expensive sense amplifier, but it is
attractive for associative memories for three reasons:
1) A fast read capability
2) A non-destructive read-out feature
3) A bipolar sense signal
Cylindrical thin film plated wires are also being investigated for
associative memory applications. The cylindrical thin film plated wire
appears attractive in that it offers the possibility of a large sense signal
resulting from the closed magnetic path, and requires a relatively small
read write drive current.
2-141
The work at Bunker-Ramo Corporation on an associative memory under a
Navy Bureau of Ships contract illustrates the Biax approach to associative
4
memories. This associative memory will store 2048 words of 36 bits each.
It will have the ability to search the entire memory for an exact match, for
gieater than, for less than, for between limits, and for simultaneous or
successive combinations of these. It will also piovide for arbitrary masking
to control the bit positions on which the search is based. The results of the
search can indicate match or no match, a count of the number of locations
upon which a match was obtained, the addresses of all positions In which a
match was obtained, or the data contained in each location in which a match
was obtained. Optionally, all matched words can be modified automatically
after a search. The basic search rate is 100 nanoseconds per bit position
for an exact match. Therefore, a search of all 26 bit positions of the
entire memory would require 3.6 microseconds. For all other search modes,
300 nonoseconds per bit position are requ i red-- 10 . 8 microseconds to search
on all 36 bits. As a conventional read-write memory, the cycle time is 6
mi croseconds.
One hole of the Biax memory is wired as a conventional coincident
current read/write memory, using the Biax element in the same way that a
simple magnetic core would be used. The orthogonal hole is then used for
the associative feature. In the associative mode, given bit positions
of all words are interrogated simultaneously, and successive bit positions
of the memory are interrogated sequentially. All bits of a given word are
sensed by the same sense wire, so that a sequential signal corresponding
to the sequencing through the bit positions is obtained from an individual
sense amplifier for each wotd position. The interrogation is conducted so
that a negative signal from the bipolar sense signal during the active part
of the cycle indicates the absence of a match. Therefore, the indication
of any negative signals from a particular sense amplifier during this active
period indicates a match for equality in that word position. The determina-
tion of greater than or less than is accomplished by noting the bit position
from which the first negative signal occurs for a particular word. This
signal, indicating a difference in that bit position, combined with the
indication of whether the control word has a one or a zero in that bit
position will then indicate which is greater.
2-]42
2.5.5.5 Planar Magnetic Thin Film Memories
In most of the approaches to planar magnetic thin film memories,
an array of discrete elements is fabricated by vacuum deposition processes
C C "7 O
on a substrate such as glass. ' ' ' Each element is about 1,000 angstroms
thick and is approximately 25 x 50 mils In area. The X and Y drive lines
are either vacuum deposited on the same substrate with appropriate insula-
tion between them or, in some cases, the drive lines and sense lines are
fabricated on a separate substrate which is then mechanically superimposed
over the one containing the magnetic elements.
Most memories of this type depend upon a predetermined orientation of
the magnetic domain based on the use of anisotropic material to provide an
"easy access" in one direction. The electrical signals then attempt to
rotate the direction of orientation of the magnetic domains. The fluxes
resulting from the electrical signals leave the orientation of the magnetic
domains in other than their normal condition. This type memory can be read
non-destruct ively by disturbing the orientation of the magnetic domains which,
when the disturbing signal is removed, return to their previous condition,
due to the anisotropic material.
Since this type of magnetic memory does not use a closed flux path,
very low sense signals are generated. The greatest difficulty with this
type of memory has been in achieving adequate uniformity in the individual
elements deposited on a plane. Uniformity is a particularly serious problem
since it depends upon the magnetic orientation of the deposited material
as well as upon the physical uniformity of the element. The lack of a
closed flux path further contributes to the problems resulting from lack
of uniformity. Because of these problems, this type of memory is usually
operated in a work-oriented fashion rather than a coincident current mode.
This increases the cost of the associated electronics. On the other hand,
the use of domain wall rotation and the absence of a closed flux path permit
higher speed operation In this type of memory than in a magnetic core matrix
memory .
2-143
Many companies are working on magnetic thin-film memories. It has been
predicted that within the next few years they will replace magnetic core
matrix memories for those high speed applications (less than 1 microsecond
cycle time) in which 1 i nea r- select core memoreis are now used. However, it
is considered unlikely that they will replace coincident current magnetic core
matrix memoreis for large capacity, slower speed applications (cycle times of
2 microseconds or greater) .
Thin-film elements can now be rotated in approximately one nanosecond,
but the associated electronics cause selection to take at least 35 nanoseconds.
Thin-film memories can provide non-destructive read out, require less power
than core memories, and are batch fabricated. However, the low sense signal
level requires a good differential amplifier; fast rise time circuits are
required, and it is difficult to adequately control the magnetic variables to
produce uniform elements. Memories of 4000 words capacity with 100 nanosecond
cycle times, and 64,000 word memories with 500 nanosecond cycle times appear
reasonable for the future.
The 400 nanosecond cycle time thin-film memory now being made by
Fabritek for Johnsville Naval Air Station is typical of the current advanced
work in thin-film memories. This memory is word-organized, contains 1024
words of 50 bits each, and operates in a 400 nanosecond cycle time. The array
is composed of film deposits approximately 900 angstroms thick that are rectan-
gular in shape, measuring 25 x 50 mi 1 . A single film is used for each bit.
Both X and Y axis lines are superimposed on the same printed circuit
card with a plastic deposit separating them. The digit lines are approximately
32 mil wide, each consisting of two 10-mil lines separated by a 12-mi 1
gap. The two 10-mil lines are joined at both ends of the board and are
terminated in a plated- through hole for interconnection. The drive line is
approximately 23 mil wide, each consisting of two 9-mI 1 lines divided by
a five-mil gap. The drive lines are also reconnected at both ends in plated
holes, which are fed through to the opposite side of the adjoining board, and
fed back parallel to the first line to provide a proper termination, and to
provide noise cancelling. Each memory plane consists of a total of 112 sub-
strates which are placed on a single printed circuit board. This is done to
2-]44
reduce the number of interconnections involved on the board and to decrease
the total number of film spots that are deposited on a single substrate thus
increasing the yield. The reverse printed circuit board is placed over the
film, and the total assembly is bolted together between two holding plates.
The drive and sense lines are then terminated.
Manufacturing costs of current film stacks are between 20(;^ and 50c a
bit, Fabritek's objective is to develop a film plane within the next year
costing approximately 5<^ a bit. The anticipated cost breakdown would be
2(^ per bit for the deposited tested film element, 1^ per bit for the overlay,
IC per bit for plane assembly and test, and ]c per bit for final stack
assembly and final test. These costs are for the fabricated planes only and
do not include electronics.
Some companies have also investigated closed flux path thin-film memories.
This approach is illustrated by the work at Philco Research Laboratory in
Newport Beach, California, where a closed flux path thin-film memory is
fabricated by first depositing a magnetic thin-film element, depositing a
conductor on top of that, and finally depositing a second thin-film element
on top of the conductor to close the flux path around the conductor. Philco
believes that this closed flux path approach will permit thin-film memories
to be operated with approximately 1/10 of the drive current (100 milliamps)
of single element planar thin films, while providing larger sense output
signals due to the complete switching of the flux. Such memories should
also be less critical to variations in the element characteristics because
of the closed flux path.
Small capacity magnetic thin-film memories have already been used in
several computers to provide a few hundred words capacity for multiple
high-speed registers requiring read/write cycle times of fractions of a
microsecond. In one machine, a small magnetic thin-film memory is used
to mechanize multiple arithmetic control registers. In another, a magnetic
thin-film memory is used to provide a small high-speed multiplexed input/output
bu"^rer that also serves as an internal scratch pad memory.
2-145
Improvements in the cost of magnetic thin-film memories are being
made rapidly, so that their cost can be expected to be competitive with
linear-select or word-oriented core memories in the near future. The choice
between linear-select magnetic core memories and magnetic thin-film memories
will be made primarily on a basis of speed vs. cost. Because of the inherent
magnetic and electronic techniques involved, there is little cost saving in
slowing down a magnetic thin-film memory to operate at speeds slower than
one microsecond cycle time. On the other hand, there c\re significant cost
penalties in trying to operate magnetic core memories at one microsecond
or less. As a result of these considerations, magnetic thin-film memories
will be in widespread use during the time scale of the planned Navy Tactical
Data System.
2.5.5.6 Cylindrical Magnetic Thin-Film Memories
Cylindrical magnetic thin-film memories are fabricated by depositing
9
a magnetic film on a wire substrate. The wire substrate then serves as one
of the electrical conductors in the system. A closed flux path is obtained
by the magnetic film surrounding the wire in a small region. A cylindrical
thin film of this type offers the advantages that the closed flux path
requires smaller digit currents and produces a larger sense signal. The
fact that the wire substrate is used as either the digit or word conductor
reduces the mechanical registration problems in the fabrication of the memory.
The major problem with this type of memory has been the difficulty of producing
satisfactory cylindrical films. Recent developments in these fabrication
techniques have made this type of memory appear more promising.
In a recent paper, "Plated Wire Magnetic Thin Film Memories,"
presented at the 1964 jntermag Conference, Danylchuk and Perneski presented
the following comparison of plated wire and planar film memories.
"1. Production method and control
Compar i son : Plated wire is produced and tested In a continuous,
serial process. Flat films are batch produced.
Conclus ion : For plated wire, small numbers of bits may be
rejected if bad. For flat films entire arrays may have to be
rejected due to failure of a single bit.
2-146
" Compar i son : The plated wires can be expected to be subjected
to strain during fabrication of a memory plane. Flat films,
which ore plated on rigid substrates, are relatively free of
strains due to handling.
Cone lus i on : During production a careful control of the
magnetostriction of the plated wire must be maintained in order
to prevent adverse strain effects.
"2. Output signal level
Compa r i son : The plated wire circumferential mode can use film
thicknesses of ly4i(10,0008) and more due to the closed flux
path at remanence. Flat film thicknesses, which are
limited by demagnetizing fields, are normally on the order of
1000^.
Conclus i on : Output signals are considerably higher for
circumferential mode plated wire memories, while bit packing
densities are comparable to those achieved with flat films. The
axial mode plated wire has no advantage over flat films in
this category since its film thickness is also limited by
demagnetizing fields.
"3. Current levels
Compar i son : Optimum coupling exists between fields produced
by currents in the plated wire and the magnetic film deposited
on the wire. For flat films, both digit and word solenoids
must be added to the film array, and the opening of these solenoids
are normally from 0.003" to 0.005". Also, slotted drive lines
are normally necessary to permit penetration of the field
generated by the uppermost conductor to the magnetic film in
the planar construction.
Conclus ion : For memory organization using plated wire as the
digit line (circumferential mode), small ( 15 ma) digit currents
are required. Alternatively, an axial mode memory which uses
the plated wire as a word line, requires relatively small ( 200
ma) word currents. On the other hand, since the plated wire
diameter (0.005") sets a lower limit on the area of the solenoid
enclosing the wire, word currents are 0.7 to 1 amp for the
circumferential mode and digit currents are approximately 0.2
amp for the axial mode. For flat films digit currents are
approximately 0.2 amp while word currents are approximately 0.3
amp. It must also be pointed out that using the plated wire
as a digit or word line leads to a much higher characteristic
impedance than for a corresponding flat film memory line."
2-147
In another type of cylindrical thin-film memory developed by NCR,
the direction of magnetization of the cylindrical elements is parallel
to the center conductor rather than circumferential to it."
In the NCR rod memory, a cylindrical thin film of magnetic material
is deposited over a conductive substrate, but the axial switching mode
produces an open flux path element. A multiple turn winding is placed
over the plated rod for each bit position. The plating material is
essentially isotropic and proper operation of the memory does not depend
upon anisotropy in the material.
NCR believes that their magnetic rod memory is more advantageous
than either magnetic core matrix memories or planar thin-film memories
in the range of 0.5 to 1.5 microseconds. They hope to lower this range
to approximately 0.3 microseconds. Magnetic core matrix memories are more
advantageous at slower speeds in the range of 2 to 5 microseconds; magnetic
thin-film memories are more advantageous at very fast speeds of about 0.1
to 0.25 microseconds. NCR estimates that in the range of 0.3 to 1.5
microseconds, the magnetic rod memory will have approximately the same
price per bit as a two microsecond coincident current magnetic core matrix
memory; approximately 25<: per bit including electronics. The electronics
are compatible with and similar to those of magnetic core memories with
a sense output of 80 millivolts being provided. NCR cites two major factors
as making the rod memory economically feasible.
1) Rod manufacture is a continuous process including
the automatic wiring of a pair of spiral wires around the
rod.
2) Tooling has been developed to permit automatic fabrication
of the memory stack (including multiple turn windings into
which the rods are inserted).
The lack of a closed flux path is recognized as a disadvantage, but the
use of a high coercivity magnetic material, permitted by the tight coupling
of the multiple turn winding, overcomes this disadvantage. As a result of
the high coercivity of the material, the tight coupling, and multiple turn
windings, fast switching and large output signals can be obtained.
2-148
2.5.5.7 Woven Screen Memory
The woven aperture screen memory, under development by Bunker-Ramo
Corporation, represents a completely different approacli to ba tch- fabr i ca t i on
12
of memory planes. In this memory, the individual memory planes ore
fabricated by a weaving process on looms similar to those used for the commercia
weaving of textiles. The woven cell is formed by weaving the proper combination
and geometry of insulated wires and bare metallic i-yires to form an orthogonal
matrix of drive and sense wires, threading magnetic cells. The magnetic cells
are formed by plating the bare metallic wires.
An electrical deposition process is used to plate a remanent magnetic
material on the square aperture cells formed by the bare metallic wires. The
plating process forms a closed flux path resulting from the plating of a pair
of bare metallic wires in the X direction and an intersecting pair in the Y
direction. The square magnetic cell formed by these four wires had been
threaded previously in the weaving process by the insulated wires representing
the sense line, the inhibit line, and the X and Y drive lines.
Since the plating process affects only the uninsulated wires, it is
possible to weave the entire plane prior to the plating process. When the
memory cell arrays are formed by the plating process, all of the sense and
drive wires are already threaded through the entire plane. The details
of the fabrication and characteristics of this type of memory were described
in greater detail by Davis and Wells in a paper, "Investigation of the Woven
Screen Memory System," presented at the 1963 FJCC.
o
A 10 bit memory assembled from 128 x 256 bit matrix planes would provide
a 10 microsecond cycle time for 36 bit words. A memory of this capacity would
be broken into 22 modules of storage planes. There is some uncertainty as to
whether a sense preamplifier would be required for each plane. This is a
question of major significance since 6336 such preamplifiers would be needed.
2-149
If the need for these preamplifiers can be avoided, the cost of a large capacity
memory of this type will be significantly reduced. This is related to the problem
discussed in connection with the laminated ferrite memory--a true low-cost
large-capacity memory requires batch fabrication and significant cost reduc-
tions in the memory electronics as well as in the storage planes themselves.
Bunker-Ramo expects the cost of coincident current woven planes in
assembled stack modules to be about 0.1 - Ic per bit in production quantities.
However, this does not include the cost of the electronics which could be
cons i derab le,
Bunker-Ramo Corporation is also working on a woven screen memory for
internal app 1 i ca t i ons--a two microsecond 8,000 word, 30 bit memory. This
memory is expected to be more rugged, to have higher temperature tolerances
and to offer a lower cost than equivalent magnetic core matrix memories. The
anticipated operating temperature range, about 105 C should be of particular
interest for military applications.
Bunker-Ramo also expects the development of integrated circuit sense
amplifiers and drivers for memories to have a significant effect on the cost
of large batch fabricated memory systems.
Pacific Semiconductor Inc. has developed an integrated circuit sense amplifier
under Air Force contract AF33 (657) - 1 1 185 . This sense amplifier is capable of
sensing a 400 microvolt input signal to provide a 1 volt output with a maximum
delay of 30 nanoseconds and a cycle rate of 20 megacycles. The write driver
amplifier represents a more difficult problem. This problem has not yet been
adequately solved.
A different approach to a woven screen memory is represented by recent work
in Japan. This has been reported by Maeda and Matsushita in a paper, "Woven
13
Thin-Film Wire Memories," and by Oshima, Futami , and Kamibayashi, in a paper
14
"The Plated Wire Memory Matrix," both presented at the 1964 Intermag
Conference, In this type of memory, plates wires are used as the storage
elements similar to those discussed previously under cylindrical thin-film
memories. The plated wire acts as the storage media, and also as the digit
drive line. In the woven array, insulated wires woven at right angles to the
2-150
plated wires act as the word drive lines. This type of woven memory differs
from the Bunker-Ramo approach in that the closed flux path magnetic element
is the plating around a single wire rather than that formed by the rectangular
intersection of four plated wires.
2.5.5.8 Laminated Ferrite Memory
One of the more promising approaches to batch fabricated memories is the
laminated ferrite memory developed at RCA Laboratories. In different versions,
this type of memory is proposed for both small-capacity, high-speed, control
memory applications, and for large-capacity, medium-speed, main internal memory
applications. Basically, the memory consists of a matrix of X and Y wires
imbedded in a solid sheet of ferrite.
In fabricating the memory plane, a pattern of parallel conductors is
printed on a glass substrate by a "silk screening type process." A film of
ferrite is spread over the conductor pattern on the substrate by a process
called "doctor-bladi ng ." After the ferrite binder dries, the ferrite is
peeled off the substrate with the conductors imbedded in the ferrite sheet.
This sheet is approximately 0.0025 inches thick. A second ferrite sheet is
made with the conductor pattern running at right angles to that in the first
sheet. A third ferrite sheet, without imbedded conductors and only 0.0005
inches thick, is inserted between the two ferrite plates with the orthogonal
conductors. This three-sheet sandwich is laminated by pressing the sheets
together at moderate pressures and temperatures. Sintering the laminated
sheets in a controlled temperature furnace provides a truly uniform isotropic
mater ia 1 .
The matrix of conductors provides the necessary wires for a two wire memory
system in which the wires in the word oriented direction are used for reading
and writing and the wires in the perpendicular direction are used for sensing
and for digit determination. The write current generates a closed flux path
in a plane perpendicular to the plane of the read/write drive wire. The
current through the digit wire rotates the magnetic vector slightly. The
methods of reading and writing in this type of array are described in detail
in a 1963 FJCC paper by Shahbender, Wentworth, Hotchkiss, and Rajchman.
2-15
RCA proposes this type of memory for two different applications.
The first is as a high-speed control memory with approximately a 100
nanoseconds read/write cycle. In this type of application, a 256 word
memory with 64 bits per word would require approximately 350 milliamps of drive
current and wou Id give a sense signal of+ 10 millivolts. The second appli-
cation is as a large-capacity medium speed internal memory with a read/write
cycle time of 1 - 3 microseconds. In this type of application, a drive
current of approximately 50 milliamps would be required to produce a sense
s i gna 1 of 1.2 mi 1 1 i vol ts .
It is necessary to operate this type of memory in a word organized
manner, but RCA does not consider this a limitation. To support this, they
state that although coincident current memories require less electronics, the
electronics are more expensive because of problems with noise in the sense
line, tolerances, and back voltage on the drive wires. On the other hand, word
organized memories require more electronics but they are less expensive since
less critical tolerances are placed on the material; there is less noise on
the sense line, and less critical tolerances are placed on the drive wire
signal. Therefore, the amount of electronics Is greater for the word
organized memory but the total cost of the electronics may not be significantly
greater because of the lesser requirements placed on them.
RCA recognizes that significant success in reducing the cost of a memory
depends upon the use of relatively cheap integrated circuits to permit the elec-
tronics associated with the memory to be made by a batch-fabrication process.
In a typical memory, the wired and tested array represents only approximately
1/2 of the total cost of the memory, with the other half going for the
electronics. As a result, techniques such as the laminated ferrite memory
alone would reduce the total memory cost by only 50% even if they were assumed
to lower the array cost to essentially zero. Hence, concentrated efforts on
reducing the cost of integrated circuit electronics for such memories are
required i f s i gn i f i cant improvements in memory costs are to be achieved.
2-152
Costs for the laminated ferrite memory are estimated to be less than
7 fi
1^ per bit for memories with capacities of 10 to 10 bits, speeds of 1-3
microseconds, and word sizes of 200-400 bits per word. RCA expects to be making
laminated ferrite memories on a commercial basis with 10 bits capacity in
7 8
two years, 10 bits capacity in four years, and 10 bits capacity in five
yea rs .
2.5.5.9 Flute Memory
Another approach to a batch fabricated memory array is the flute memory
developed by IBM. Conceptually, this memory is very similar to the RCA
laminated ferrite memory but the fabrication techniques are different. The
fabrication techniques and characteristics of this memory were described in
a paper by several authors in the April 1964 issue of the IBM Journal of Research
1 ^
and Development.
The planes in the flute memory are fabricated by sandwiching a pre-
prepared grid of wires between two dies, each of which has matching grooves
filled with a mixture of ferrite thermosetting resin binder. The grooves
containing ferrite are parallel to the word lines of the wire grid so that
when the two halves of the mold are placed together, the word line is completely
imbedded in the ferrite tube formed by the ferrite in the corresponding grooves
of the upper and lower halves of the mold. The bit lines of the wire grid are
orthogonal to the word lines and intersect each of the parallel ferrite strips
encasing the word lines,
A typical memory plane consists of 50 ferrite tubes intersected by 100
bit lines to give a capacity of 5000 bits per plane. The ferrite area
surrounding the intersection of a bit line and a word line defines an individual
bit position. Yields of 36% have been realized in the batch-fabrication
process. Cycle times of 250 nanoseconds are considered possible by IBM.
2. 153
2.5.5.10 Permalloy Sheet Memory
The permalloy sheet memory under development at Laboratory For Electronics
is another approach to batch fabrication of a low-cost large-capacity memory.
In this technique, a permalloy sheet is bonded to a substrate. A pattern is
printed on photo resist covering the sheet, and the permalloy is etched
away to leave a matrix of toroidal permalloy storage elements. Three small
holes in the center of each toroid are plated through to provide connections
from one side of the array to the other through the toroids. Printed inter-
connections on both sides of the board wire the array. This design effectively
combines a number of known techniques into a very interesting fabrication
technique for a large memory array, A paper on this memory presented by
Fuller at the 1964 Intermag Conference gives a much more detailed description
of the memory and fabrication techniques than the simplified one above.
This development program is "aimed at improving processing speeds by
providing data processing systems with random-access memories approaching
the speed of ferrite-core matrix memories, but at a cost that economically
permits mass memory capacities."
The goal in this development is a large capacity memory in which
coincident current selection is used to reduce the electronic costs, and
in which a very large memory plane (256 x 256 bits) is used to reduce
fabrication costs as well as drive, selection, and sense electronics costs.
A 6.5 million bit memory module is planned in which the smallest unit that
is individually handled in processing is a matrix plane of 65,000 bits. The
cost objectives for this development are 0,03 - O.Si per bit,
2.5.5.11 Ferroacoust i c Memory
The term ferroacoust i c memory is used to refer to two similar memory
developments being sponsored by the U.S. Army Engineering Research and
Development Laboratory at Fort Monmouth, New Jersey. These contracts are
with RCA Laboratories in Princeton, N.J. and General Dynamics/Electronics
in Rochester, New York.
2. 154
The RCA work has been desciibed in two quarterly reports submitted
to the Army ' ' The General Dynamics Electronics approach has been
20
described in Gratian and Freytag in a paper, "Ultra Sonic Approach to Data
Storage," published in the May 4, 1964 issue of Electronics.
Although this is a batch- fabr i ca ted memory approach, it differs from those
described previously in that it is not a random access memory. This type of
memory is block oriented with random access to the beginning of a block but
serial access to information within the block. A batch- fabr ica ted sol i d
state memory of this type would be a replacement for large magnetic drums,
magnetic disc files, and possibly magnetic tape.
Previous types of delay line memories suffered from the volatility
and the large amount of electronics necessitated by the requirement for
regenerating and recirculating the information in each individual delay
line. The necessity for recirculating the information results in the loss
of stored information if power is interrupted. The requirement for electronic
circuitry to be In continuous use for each individual line implies a severe
cost penalty in very large capacity memories. The fer roacoust ic approach
permits static storage of information without continuous recirculation which,
in turn, permits the electronic read and writing circuitry to be switched
from one line to another as part of the addressing and selection process.
In this type of memory, data is stored on a thin magnetic film plated on
an acoustic tube through which a center conductor is inserted. The actual
storage is in a closed flux path around the acoustic tube similar to the
storage in cylindrical thin film memories discussed previously. However,
the access is not made by coincidence of two electrical signals, as in the
memories discussed previously, but rather by the coincidence of an electrical
signal and an acoustic signal.
2-155
The concept of this type of memory is based on the change of coercivlty
of magnetic materials, such as permalloy plating, when the material is placed
under mechanical stress. An alternate approach is based on the fact that the
anisotropic characteristics of thin permalloy films are changed by stress.
In a memory based on these principles, a mechanical shock wave is
initiated in the acoustic tube by a suitable transducer (e.g. magnetos t r i ct ive
or piezo electric) . The mechanical shock wave travels down the tube at a
speed determined by the propagation constant of the material. The coercivity
of the magnetic material plated on the tube changes as the ^ock wave passes
under it and returns to its normal condition after the shock wave has passed.
As a result, a temporary change in the coercivity of the magnetic media is
propagated down the line.
An electrical signal, corresponding to the serial bit pattern of I's
and O's to be stored, is placed on the center conductor. This signal causes
the magnetic cylinder to be linked by a varying flux pattern depending upon
the bit pattern represented by the electrical signal. The flux pattern
generated by the electrical signal corresponding to a "1" is sufficient to
change the state of magnetization of the magnetic material in an area where
the coercivity has been altered by the mechanical stress wave. The magnetic
flux corresponding to a "0" signal will not alter the state of magnetization,
even in those areas where the coercivity has been changed. As a result, a
bit position is determined by the time coincidence of the electrical signal
and the acoustic signal. As the acoustic signal travels down the line, a
'" 1" or "0" corresponding to the electrical signal will be written in the
position defined by the front of the shock wave.
The time required to read a complete block corresponds to the time
the shock wave requires to travel from one end cf the line to the other.
The rate at which the stored information is read or written corresponds to
the frequency of the electrical signal representing the bit pattern. This
frequency is limited by the physical size of the area of magnetic material
whose coercivity is altered by the shock wave at a given instant. This
in turn determines the bit density and hence the capacity of a line of a
given length.
2-156
It is Important to note that the information is stored statically in
the magnetic plane and is not carried by the acoustic wave. This is in
contrast to a normal acoustic delay line in which the information is actually
carried by the acoustic wave. In the f er roacous t i c type memory described
here, the acoustic signal acts only as an access mechanism. A single
acoustic signal is required to read a complete line or write a complete line.
Read-back is achieved by transmitting an acoustic signal down the line and
sensing the changes in the electrical signal generated on the center con-
ductor by the magnetos tri ct ive effect of the shock wave traveling down the
line. This signal varies, depending upon the stored pattern on the line.
If this fer roacoust i c approach proves feasible, it offers the following
advantages :
1) Static storage
2) Semi-serial access not requiring a physical coincidence of selection
wires for each bit position.
3) Ability to switch read and write mechanisms from one line to
another.
4) Large capacity semi- random access bulk storage with no
mechanically moving parts.
5) Possibility of off-line storage by plugging alternate blocks
of delay lines into a read/write device.
5) Low cost per bit of storage.
The goals of the Army development program are to provide a block oriented
random access memory with the following characteristics:
4,000,000 characters per module
4096 blocks per module
1024 blocks per module
approximately 0.002(J per bit
(off-line storage cost not including read/write electronics)
1 microsecond access to a block,
several megacycle read/write rate.
If this development proves feasible, it will be of great significance
to future computer systems. At present, there is no economic all electronic/
mechanical replacement for large capacity electro-mechanical storage devices
such as magnetic disc files, magnetic card memories, and magnetic tape units.
2-157
Such a replacement will be essential to future systems if the speed, cost,
reliability and size limitations of electro-mechanical input/output and
auxiliary storage equipments ave to be avoided for very large capacity
storage functions (now handled by devices such as disc files and magnetic
tapes) .
A future all electronic magnetic replacement will almost certainly be
a block-oriented rather than a random-access- type storage device. Although
a large capacity, random access, mass memory offers certain unique advantages, it
is very unlikely that such a device can compete on a cost per bit basis with
semi-serial electromechanical devices. The requirement for intersection of
electrical signal lines for each bit position, and the excess electronics will
not provide on-line storage costs of a few millicents per bit by 1970. A
semi-serial or block-oriented device providing random access to a block of
information, but serial access within the block will be necessary to permit
the read/write electronics to be time shared by a serial bit train.
2.5,5.12 On-Line Auxiliary Storage
On-line auxiliary storages are frequently referred to as mass memories.
On-line devices of this type are used to provide a large-capacity, fast,
semi- random-access storage. They should be under direct on-line control of
the computer, addressable by the computer, eraseable, and reuseable. All
devices of this type that are currently available are electromechanical.
All electronic/magnetic on-line auxiliary storages for military applications
will be available by 1970 but all of these (with the possible exception of
the ferroacousti c storage discussed previously), will be significantly more
expensive in terms of cost-per-bit for very large capacity storage. This
results largely from the fact that the all-electronic/magnetic approaches
(e.g. laminated ferrite memory, woven screen memory, flute memory, etc.)
require addressing each individual word. The electromechanical devices
are block oriented in the sense that access is made to a particular track on
a disc, drum or card, and then all information stored In that track (or
block) is read or written serially by the same electronic circuitry.
2-158
Of the all-electronic/magnetic approaches to large capacity memory that
have been discussed in previous parts of this report, only the ferroacous t i c
approach is block oriented and hence offers some promise of competing v/ith
the electromechanical devices by 1970 on a purely economic basis.
As a result of previous studies by one member of the study team,
more detailed Information is available at this time on electromechanical
2 1 22
mass memories ' The information presented here is indicative of the
type of critical detailed evaluation and comparisons that will be made for
the other memory areas during the remainder of the study. The characteristics
of the major types of electromechanical mass memories are summarized in the
table shown in Table 2.-4.The values shown were chosen as typical of each type
of unit but frequently they represent a wide range of possible values. In
some cases, certain characteristics of an individual device may vary signifi-
cantly from the values shown. The characteristics of primary interest include
capacity, cost, average access time, and data transfer rate. Some of these
are difficult to compare because of the different physical characteristics
of the devices. For example, a large magnetic drum with a head for every
track will have a continuous data transfer rate equal to the instantaneous
transfer rate if the heads are switched electronically. However, the con-
tinuous data transfer rate for a disc file with moving heads will be
significantly greater than the instantaneous transfer rate due to the
necessity for interrupting data transfer while moving the head from one
position to the next. Similar differences on a more detailed level exist
between different devices of the same type.
In comparing costs, a detailed investigation is usually required to
determine whether prices quoted for different units Include comparable elect-
ronics (I.e., controllers, buffers, switching, amplifiers, etc.). The
estimated costs shown In the table are user's costs (rather than manufacturing
costs) and assume a moderate amount of associated electronics.
TYPE OF
DEVICE
Large
Fixed-Head
Mag. Drums
ON-LINE
CAPACITY
PER-UNIT
IN CHAR.
0.2 X 10
to
5.0 X 10
TYPICAL
ON-LINE
COSTS IN
(i/CHAR.
AVERAGE
ACCESS
TIME
DATA
TRANSFER
RATE IN
CH/SEC.
2.0
15 ms
100,000
to
200,000
REMARKS
Mov i ng-Head
Magnet i c
Drums
4.0 X 10
65 X 10
0.3
100 ms
50,000
to
50,000
Fixed-Head
Magnet i c
Di sc Fi les
10 X 10
to
40 X 10
0.5
20 ms
100,000
to
350,000
Offers Promise for
Mi 1 i ta r i zat i on
1 Dimension
Moving-Head
Mag. Disc.
10 X 10
^° (
250 X 10
0.2
100 ms
100,000
to
400,000
Contract has been awarded
for militarizing one device
of this type.
2 Dimensi on
Mov i ng-Head
Mag. Disc
10 X 10
150 X 10
0. 15
500 ms
50,000
to
100,000
Relatively obsolete
Removable-
Stack
Di sc Fi les
2.0 X 10
to
7.0 X 10
0.5
(on- 1 i ne)
0.01
(off- 1 i ne)
150 ms
160,000
Off- 1 i ne storage
capaci 1 i ty
Magnetic
Card
Fi les
5.5 X 10
to
340 X 10
0.03
(on- 1 i ne)
0.0001
(off-line)
250
100,000
Off-line storage capability
Contract has been awarded
for militarizing one device
of this type
TABLE 2-4 SUMMARY OF CHARACTERISTICS OF ELECTROMECHANICAL MASS MEMORIES
to
2-160
Access time offers another illustration of the difficulties of
comparing different types of mass memories. It is difficult to compare
the access times even for different devices of the same type--for example,
different designs and makes of disc files. It is considerably more difficult
to compare the access times for completely different types of mass memories
due to differences in the methods of making mechanical access since the
total mechanical access is usually made up of a number of separate components.
As a result of problems of this type, comparison tables such as the one
shown in Table 2- 4 present at best a gross comparison. In selecting a device
for any specific application, it is necessary to go into a more detailed com-
parison of the specific pecularities and quirks of each of the leading
contenders as they relate to that application if a proper decision is to be
made.
1) Large Magnetic Drums
Until recently, the capacity of large magnetic drums ranged from
approximately 200,000 to 1,000,000 characters per unit for those
with fixed heads, and approximately 4 to 10 million characters for those
with moving heads. However, one manufacturer recently announced a large
dual drum unit with moving heads providing a capacity of 65 million
alphanumeric characters. In this unit, two very long drums (over six
feet) are rotated on parallel centers with the surfaces close enough
to one another to permit a single access mechanism to position sets of
64 heads--32 on each drum.
The choice between these two types of drums depends largely upon
whether access time or capacity Is the more important consideration.
The fixed-head drum also implies a higher cost per bit of storage due to
the number of heads and the switching circuitry required.
2) Magnetic Disc Files
A magnetic disc file consists of a stack of disks (usually 5 to 100)
rotating on a common shaft. The discs are usually between 1 12 and 3
feet in diameter. Magnetic disc files can be classified as those with
fixed heads (one head per track on each disc) , those with moving heads,
and those with removable disc stacks (and moving heads). Disc files
with moving heads can further be divided into those in which the heads
move in one dimension only (in and out among the stack of discs) and
those in which the heads move in two dimensions (up and down the stack of
discs as well as in and out among the stack). The major effects that
these differences have on the characteristics of the devices are indicated
in the Table too.
2-161
In general, the larger the number of bits that can be accessed by a
single head and selection mechanism, the lower the cost per bit and the longer
the access time.
a). Fixed-Head Magnetic Disc Files
Disc file storage units with fixed heads usually involve a limited
number of discs, a maximum number of bits per track, and a fixed head
for each track. This type of storage permits a higher track density
since the fixed heads eliminate the need for mechanical positioning of
the head and the resulting allowances for mechanical tolerances. The
large multiplicity of heads, and the required electronic switching
between heads, results in a significantly higher cost per bit than
for the moving-head type disc storage.
Although this type of disc storage is somewhat similar in functions
and characteristics to fixed-head magnetic drums, the use of three
dimensions instead of two permits greater volumetric ef f i c i ency--grea ter
storage capacity in a more compact unit. There are, of course, also differ-
ences in the mechanical design problems between disc and drum units, but
these are outside the scope of this paper.
b) . Moving-Head Magnetic Disc Files
The first commercially available magnetic disc files involved a
two-dimensional head movement. A single head mechanism was moved up
and down parallel to the disc stack and shaft to select one of a
number of discs, and then moved in between adjacent discs to select
the desired track. In this unit, the head-mount arm straddled a disc
to provide a head to read the upper surface of the disc and another
head to read the lower surface.
Although some modern large capacity disc files also operate on
this principle, most of the present units involve a one-dimensional
movement. A head mount is inserted between each pair of adjacent
discs, usually with one head reading the lower surface of the upper
disc and another head on the same mount reading the upper surface of
the lower disc. This type of disc file provides a much faster access
by eliminating the necessity for moving the heads in the dimension
parallel to the disc shaft. The penalty paid for this faster access
is the increase in the cost per bit of storage due to the cost of
the larger number of heads and the electronic switching between
heads compared with the cost of a disc selector mechanism.
The one-dimensional movement permits two secondary advantages that
are not apparent from the comparisons in the Table. Since there is at
least one head for each disk, it is possible to provide a larger number
of read and write amplifiers to permit reading or writing multiple tracks
simultaneously with a significant increase in the effective instantaneous
data transfer rate.
2-162
Even without simultaneous reading or writing from all heads, the
ability to switch electronically between heads increases the information
that can be transferred without moving the arm. Appropriate organization
of the problem can reduce the number of arm movements required, thereb/
increasing the effective speed of operation.
The insertion of the set of head mounts between pairs of adjacent
discs can be, and has beed, accomplished in several different ways
mechanically. In one design, the head mounts for all discs are moved
together by a common track selection mechanism. As a result, all of the
heads are moved In and out simultaneously to corresponding tracks on each
disc. This can be pictured as a comb of head mounts moving in and out
perpendicular to the disc shaft. For any one position, the tracks being
read or written on each of the discs describes a cylinder conceptually
similar to a magnetic drum with wide track spacing. Another design
provides independent head positioning mechanisms for each disc. If
utilized, the ability to Independently access tracks on different discs
can permit a significant decrease in effective access time since
several accesses to different discs can be overlapped or performed
s imu 1 taneous ly .
c) . Removable-Stack Disc File
The newest addition to the disc storage family is the removable-
disc-stack unit. In this device, a drive mechanism handles a small
stack of discs that can be removed, replaced, and interchanged with
other stacks.
This device provides a compromise between the off-line storage
capability of magnetic tape, and the on-line fast access capability
of larger mass memories. A series of disc stacks can be stored away on
a shelf and put on the drive mechanism as required. Each disc stack has
approximately one fourth the capacity of a tape reel with ten times
higher cost. However, all data within a stack can be on-line and
addressable by the computer to provide fast access within blocks of
two million characters at a time. This is particularly well suited to
the requirements of many types of business problems for large total file
storage capability but on-line fast access to only a segment of this In
any given processing operation.
3) Magnetic Cards
The magnetic-card type of mass memory, which preceded the removable-
stack disc storage by over two years, is quite different physically and
mechanically. However, from a systems and applications standpoint, the
two are somewhat similar in that the magnetic card memory also provides
a certain amount of on-line storage capacity and an almost limitless
amount of off-line storage capacity. The magnetic-card type offers
an advantage over disc files in that Individual cards can be copied,
inserted, removed, or replaced.
2-163
In one device of this type, oxi de-coa ted MyKir Ccirds, approximately
3 X 14 inches in size, are hung from rods in the magazine. These rods
may be selectively turned to select the card with binary-coded notches
corresponding to the rods that have been turned, thus providirig the
ability to select any card from the magazine at random. The selected
c^i rd is then dropped to the surface of a rotating drum cind accelerated
to the surface speed of the drum so that it can be read or written
while passing under a set of heads. The card may be held on the drum
for rereading or for reading another set of tracks on the same card.
When it is released from the drum, it is automatically returned to the
magazine. Its location in the magazine is immaterial since the selection
is by the coded notches in the card and the combination of rods that
are turned rather than by physical location.
The major advantages and disadvantages of the different types of
mass memories are summarized in Table 2-5.
TABLE 2-5
ADVANTAGES AND DISADVANTAGES OF DIFFERENT MASS MEMORIES
2-164
TYPE OF MASS MEMORY
Fixed-Head Magnetic Drums
Moving-Head Magnetic Drums
Fixed-Head Magnetic Discs
Two- D I mens i on Moving- Head
Magnetic Discs
One- D i mens i on Moving-Head
Magnetic Discs
Removable-Stack Discs
Magnetic Card Memory
ADVANTAGES
Fast access, no mechanical
head motion, high continuous
data transfer rate.
DISADVANTAGES
Low capacity, high cost
per bit, poorer volu-
metric efficiency, large
number of heads.
Large capacity, low cost Poorer volumetric
per bit, possibility of efficiency, relatively
parallel reading or writing large number of heads
from multiple heads to greatly for medium speed access
increase instantaneous data or slower access if
transfer rate. fewer heads.
Fast access, medium capacity. High cost per bit of
no mechanical head motion, storage, large elect-
high continuous data trans- ronic switching matrix,
fer rate. large number of heads.
Large capacity, minimum
number of heads, low cost
per bit
Large capacity, possibility
of multiple simultaneous
accesses if heads are posi-
tioned independently, low
cost per bit compared to
fixed head units, possibility
of parallel reading or writing
from multiple heads to greatly
increase instantaneous data
transfer rate.
Large off-line capacity, low
cost per bit of off-line
storage, combines on-line
random-access capability
with large off-line capacity.
Large off-line capacity, low
cost per bit of off-line
storage, combines on-line
random-access capability
with large off-line capacity,
Individual cards can be
copied, replaced or inserted.
More complex position-
ing mechanism, slowest
access, slow continuous
data transfer rate.
Relatively large number
of heads, somewhat hlghe
cost per bit compared tc
two-dimension disc unit,
medium speed access.
Limi ted on- 1 i ne
capacity, higher cost
per bit of on- 1 i ne
storage.
Slower access, card
wear and replacement
necessitates eventua
rewriting of entire
card.
2-165
Future improvements can be anticipated. In 1962, A. S. Hoagland
pointed out that the storage density of one manufacturer's commercial
disc files increased from 2000 bits per square inch in 1956 to 25,000
bits per square inch in 1961, He then predicted that storage densities
of "one million bits per square inch (e.g. approximately 5000 bpi, 200 Lpi)
will become the state-of-the-art" within the next few years. A few months
earlier, M. Jacoby predicted densities of 3000 bpi and 500 tpi (1.5 million
bits per square inch) would "become common-place in a few years". He then
indicated that these densities could provide storage capacities of 10 to 100
billion bits if a possible increase in the physical size is also considered.
Thus, increases in capacity of tens to hundreds of times over the largest
present mass memories can be anticipated.
The cost per unit can be expected to decrease even with the larger
capacities as the technology is improved and more manufacturing experience
is obtained. Hence, the cost per bit of storage can also be expected to
decrease by one to two orders of magni tude--poss i b ly to 0.0001 cents per bit
for the mass memory itself (not incl ud i ng control and buffering electronics).
Although the picture for the future of capacity and cost appear bright, there
is little hope for significant improvements in average access times for moving-
media mass memories. Due to the inherent mechanical motions involved, we cannot
expect improvements of as much as an order of magnitude over available devices.
For significant improvements in access time, we must turn to the non-moving-
media type devices.
It is likely that continued improvements and innovations in moving-
media mass memories will provide ultimate capacities, access times, data trans-
fer rates, and costs superior to those indicated above. Just as the development
of the floating head permitted densities and rates in excess of those previously
anticipated for fixed heads, unforeseen developments may well serve to push
the limits of these devices beyond expectations. An example of work on one
such development has been described in Hoagland. This is a disc unit in which
the head in positioned on a track under control of a servo system with the
signal read from the track being part of the control loop to permit far greater
track density and multiple access arms.
2-166
2.5.6 Memory Availability in the 1970-80 Period
A preliminary estimate of characteristics expected to be available,
feasible, and competitive for use in a 1970 system are shown in Tables
2-6, 2,7, and 2.8 for registers and high-speed control memories,
main high-speed internal memories, and on-line auxiliary storage. Only all-
electronic/magnetic technologies are shown. There is not sufficient basis
at this time of anticipate an all electronic/magnetic off-line auxiliary
storage. The ferroacous t i c storage discussed previously appears to be the
most likely candidate, but it is too early in the development of this device
to determine whether it will be feasible by 1970.
The comparisons in the three tables do not include memories that are
expected to be obsolete by 1970 and memory techniques that appear promising
on a longer time scale but are not expected to be operational by 1970. The
characteristics shown for different types of memories are those expected to
be realizable for a memory to be operational in 1970. These preliminary
comparisons will be refined during the remainder of the study and additional
memory technologies will be added to the comparisons.
TYPE OF
STORAGE
TYPICAL TYPE R/W
CAPACITY ACCESS CYCLE
(WORDS) TIME
R/W VOLATILE POSSIBLE
RATE DATE OF
1st PROD,
BATCH
FABRICATION
TECHNIQUE
Integrated Ckt
FF Registers
25
Random 125 ms 40 mc
Yes
1966
Diffusion and Vacuum
Depos i t i on
Planor Thin-
Fi Im
512
Random 100 ms 10 mc
No
965
Multi-Layer Vacuum
Depos i t ion
Cy 1 i ndr lea 1
Thi n-Fi Im
(Magnetic-Rod)
512
Random 250 ms 4 mc
No
1965
Plating and Automatic
Cost Winders
Laminated Ferrite
512
Randon 100 ms 10 mc
No.
1966
Silk Screen i ng
"Doctor-Blading" ,
Lami na t i on
Linear Select
Magnetic Core Matrix
512
Random 350 ms 3 mc
No
966
Tunnel Diode
512
Random 10 ms 100 mc Yes
1965
None
Flute
512
Random 250 ms 4 mc
No
1966
Ferr i te Mol d i ng
TABLE 2-6
Preliminary Estimate of Characteristics of Registers and High-Speed Control Memories In 1970
I
TYPE OF
STORAGE
TYPICAL
TYPE
R/W
R/W
VOLATILE
POSSIBLE
BATCH
CAPACITY
ACCESS
CYCLE
Rate
DATE OF
FABRICATION
(WORDS)
TIME
(WORDS)
1st PROD.
TECHNIQUE
Conti nuous-Sheet
Cryogeni c
2 X 10
5
Random 1 .0 ms 1 mc
1969
Multi- layer Vacuum
Depos I t i on
Laminated Ferrite
Permal Toy- Sheet
0.2 X 10 Random 2.0 ms 0.5 mc
No
1968
S i 1 k- Screen i ng
"Doctor-Bladl ng" ,
Lami nat i ng
Woven- Screen
0.2 X 10^
Random
2.0 ms
. 5 mc
No
1966
Weaving, Plating
Magnetic Thin-FI
1m
0.2 X 10^
Random
0.5 ms
2 mc
No.
1967
Vacuum Depos i t i on
2.0 X 10 Random 10.0 ms 0.1 mc
No
1967
Silk-Screening, Etching,
Plating
Flute
0.1 X 10 Random 2.0 ms 0.5 mc
No
1968
Ferr i te Moldi ng
Linear Select
Magnetic Core
Ma t r i X
0.2 X 10 Random 1.0 ms 1 mc
No
1965
Glass Delay Line
0.02 X 10
Serial
Random
20.0 ms 1 mc
Yes
1965
Glass Rod
TABLE 2-7
Preliminary Estimate of Characteristics of Main High-Speed Internal Memories In 1970.
I
00
TYPE OF
STORAGE
TYPICAL
CAPACITY
(WORDS)
TYPE
ACCESS
ACCESS
TIME
R/W
CYCLE
TIME
R/W
RATE
VOLATILE
POSSIBLE
DATE OF
1st PROD.
BATCH
FABRICATION
TECHNIQUE
Conti nuous- Sheet
Cryogeni c
20 X 10^
Random
—
5.0 ms
0.2 mc
?
1970
Mu 1 1 i - Layer
Vacuum
Depos i t i on
Woven-Screen
5 X 10^
Random
-
10.0 ms
0. 1 mc
No
1968
Weav i ng ,
Plati ng
Permal loy- Sheet
5 X 10^
Random
-
100.0 ms
0.01 mc
No
1958
S I 1 k-Screen i ng ,
Etchi ng,
Platl ng
Ferro-Acoustic
20 X 10^
Serial/
Random
1 ms
(to bloc
:k)
-
3 mc
No
1969
Plating,
Acoustic
Cy 1 inder
TABLE 2-8
Preliminary Estimate of Characteristics of On-line Auxiliary Storage in 1970
2-180
During the time from 1970 to 1980, continual improvements will be
made in the characteristics of most of these memory types. Other types of
memories not feasible for a 1970 system will probably be developed to a point
that they can be included in an operational system prior to 1980. These
may include low-cost, large-capacity random access mass memories, integrated
semiconductor memories, electron spin echo storage, large-capacity low-cost
magnetic film domain wall storage, and perhaps high resolution electron-beam
fabricated storage systems. There will undoubtedly be some radically new
memory techniques developed during the 1970-80 period, but a large part of
the improvement in memory characteristics and capabilities during that time
period will result from continued improvements of memories now in use or
under development. This is particularly true with respect to improved
batch-fabrication techniques. Although there is a possibility that some exotic
new memory techniques, such as a high-speed random access read/write memory
based on the use of lasers, will be developed during that time period, it is
certain that new methods of fabricating magnetic memories will be developed
that will have significant and dramatic effects on the cost, speed and
capaci ty .
2.5.7 Limitations of Present and Planned Memories
It is difficult to place ultimate limits on the cost, speed, and capacity
of different memory types since the violation of basic physical laws has not
been the limiting factor to date and probably will not be for the foreseeable
future. The comparison tables given in Section 2.5.5 indicate the characteristics
anticipated for a 1970 system but these are not ultimate limitations in most
cases. It is important to note that in considering the limitations, the set
of characteristics must be taken as a whole. For example, for a particular
type of memory, a certain combination of speed and cost may be anticipated
for a 1970 system. However, if the capacity were decreased significantly, the
speed could be increased and if the speed were decreased, the capacity could
be increased. Therefore, the characteristics shown in the tables in Section
2.5,5 should not be considered as limitations on any individual characteristic
2-18]
but rather as a reasonable expectation for characteristics for 191^,
A number of memory experts have given consideration to ultimate limitations
of memory technologies. One of these is Dr. J. A. Rajchman of RCA Laboratories
who prepared the diagram shown in Figuie 2-35. This diagram shows Dr. Rajchman' s
estimate of the limitations in terms of speed vs. capacity for different
memory technologies. The difficulty of placing such limitations on a
rapidly developing technology is indicated by the fact that the diagram shown
in the figure is the latest of a number of similar diagrams prepared by
Dr. Rajchman over the past two or three years each being updated to reflect
changes in the technology since the previous one.
2-182
100/xS lO/xS
/jls lOons
CYCLE TIME
Ons
Figure 2-35 - Storage Capacity and Cycle Time of Memories
(After Rajchman)
2-183
2.5.8 Recommended Developments to Meet ANTACCS Needs
No development efforts are necessary to assure availability of
memories meeting ANTACCS needs in 1970 from a performance stand-point.
However, development efforts will be required to assure that certain
types of these memories meet the operational requirements for temperature,
shock and vibration, nuclear radiation effects, electro-magnetic interference
requirements, humidity, and perhaps reliability and maintainability. It is
anticipated that the solid state electronic and magnetic memories will meet
ANTACCS requirements for reliability and maintainability but if electro-
mechanical memories such as magnetic drums, magnetic disc files, and magnetic
card memories have not been completely replaced by all electronic/magnetic
memories by that time, additional development efforts will be required to
further militarize these electromechanical devices. Some work in this
direction is currently underway but additional work will be necessary to
meet fully the requirements of a 1970 ANTACCS System. The requirements
for an ANTACCS for that time period are probably more stringent and would
necessitate development work directed toward reducing the size and weight
of these devices as well as meeting the environmental conditions discussed
above.
This is considered to be an interim effort since it is believed
that electromechanical devices of this type will ultimately be replaced
by all electronic or magnetic memories. However, at this time it does not
appear safe to assume that this will be accomplished in all cases in time
to satisfy the requirements of the 1970 ANTACCS system. Hence, it will
be necessary to further improve the environmental characteristics, reliability,
and maintainability of present types of electromechanical memories (particularly
disc files and magnetic card memories) for these systems.
2-184
2.5.9 Evaluation Criteria Recommended
In evaluating different types of memories for use in a 1970 ANTACCS
System, it will be necessary to consider a relatively large number of
characteristics and parameters. However, many of these can be quickly
considered for most types of memories and will not require an elaborate
and detailed comparison. These characteristics will be noted for particular
memories only where they significantly increase the desirability or appear
to be a major deterrent. to the use of a particular type of memory. The
parameters and characteristics to be considered in comparing and evaluating
different types of memories will include:
Type of storage media or phenomena involved
Access time
Type of access
Read/write cycle time
Read-only cycle time
Read/write rate
Device switching time
Storage capaci ty
Storage density
Static or dynamic
Eraseable
Non-destructive read out
Volatile or non-volatile
Addressing and selecting techniques
Size and weight
Operating temperature range
Susceptibility to shock and vibration
Susceptibility to nuclear radiation effects
Susceptibility to electromagnetic interference
Generation of electromagnetic interference
Susceptibility to humidity
Other environmental considerations
Approximate or relative cost per bit
Range of reasonable memory system costs
Rel iabi 1 i ty
Ma inta inabi 1 i ty
Logistics Requirements
Batch-fabrication techniques
Fabrication and packaging problems
Stand-by power requirements
Operating power requirements
Approximate or estimated date of first production quantity
appl i cations
Special features (e.g. associative)
Functional uses in a computer or digital system
Special requirements (e.g. cooling or refrigeration)
2-185
Although all of the above characteristics should be considered, it
will usually be necessary to make a detailed comparison and evaluation only
oni the basis of the following characteristics: type of storage media or
phenomena, access time, type of access, read/write cycle time (for random
access memories), read-only cycle time (for read-only memories), storage
capacity, volatile or non-volatile, eraseable, relative cost per bit,
adaptability to batch-fabrication techniques. The effect or implication
of other characteristics will be noted where appropriate for particular memories
as "remarks." For example, the operating temperature range will be noted only
where it is extremely difficult for a particular type of memory to meet the
temperature ranges required in the military environment.
2.5.10 Conclusions and Recommendations
Memories in all categories for use in a 1970 ANTACCS System should
and can be manufactured by various batch-fabr icat ion techniques. Such
techniques for fabricating large memory arrays as units rather than by the
assembly of large numbers of discrete elements, are well along at this
time and are receiving considerable attention from the industry. The memory
function Is particularly adaptable to batch-fabrication techniques since
it consists of large numbers of similar elements or circuits and hence is
highly repetitive. This is true on one extreme for small high speed one word
registers that might be fabricated as a single integrated circuit array, and on
the other extreme for very large capacity on-line auxiliary memories such as
a cryogenic memory, laminated ferrite memory, or woven screen memory. It
is certain that techniques of this type will be feasible, economic, highly
developed and in widespread use by 1970.
However, it is not certain that large capacity on-line auxiliary
memories of this type will be competitive with electromechanical memories
(e.g. magnetic disc files and magnetic card memories) on a cost-per-bit
basis by that time. It may be necessary to use some electromechanical
device for this mass memory function, or to recognize and accept a cost
penalty for using an all electronic or magnetic approach. This is particularly
9
true for mass memories with capacities of 10 bits and above. Although these
2-186
very large capacities can be achieved by using multiple banks of smaller memories,
this will probably not be economically feasible. As a result, moving magnetic-
media type electromechanical type memories will likely be used for these
requ i rements .
Associative memories will be feasible and available but their use
will depend upon developments in machine organization and upon significant
cost reductions. Relatively small capacity associative memories will probably
be used in conjunction with the main high-speed random-access internal memory for
functions such as indexing and perhaps some list processing. However, it is not
believed that large capacity associative memories serving as the main internal
memory will be economically justifiable or feasible by 1970.
Conclusions concerning the relative advantages and disadvantages of
specific types of memories will be developed during subsequent parts of this
study and recommendations will be made as to specific types of memories to use
in different subsystems and for different functions.
2-187
References; Memories, Section 2.5
1. "A Large Capacity Cryoelectric Memory with Cavity Sensing."
Burns, L. L., Chr i st iansen , D. A., and Gange, R. A., Proceedings 1963
FJCC, pp 91-99, November 12-14, 1963.
2. "A Cryogenic Data Addressed Memory," Newhouse, V. 1., and Fruin, R. E.,
Proceedings Spring Joint Computer Conference, Vol. 21, pp 89-100,
May 1-3, 1962
3. "Design of a Fully Associative Cryogenic Data Processor," Pritchard, J. P.,
Jr., and Wald, L. D. , 1964 Proceedings of Intermag Conference, pp 2-5-1 -
2-5-4, April 1964.
4. "Theory, Organization, and Performance of a Search Memory," Koerner,
R. J., and Searbrough, A. D. , Local Symposium on Search Memory,
Los Angeles District of IEEE, May 26, 1964
5. "Magnetic Films - Revolution in Computer Memories," Chang, C. and
Fedde, G., Proceedings 1962 FJCC pp. 213-224, May 1962.
6. "The Future of Thin Magnetic Films", Bittman, E. E., Large Capacity
Memory Techniques for Computing Systems , pp 411-420, Macmillan
Publishing, New York, 1962.
7. "High Density Magnetic Film Memory Techniques," Crowther, T. S., 1964
Proceedings Intermag Conference, pp 5-7-1 - 5-7-6, April, 1964
8. "Future Developments in Large Magnetic Film Memories," Raffell, R. I.
Ninth Annual Conference on Magnetism and Magnetic Materials, Atlantic
City, N. J., November 1963.
9. Journal of Applied Physics 30, Long, T. R. , pp. 1235, 1960.
10. "Plated Wire Magnetic Film Memories," Danylchuk, I and Perneski , A. J.,
1964 Proceedings Intermag Conference, 5-4-1 - 5-4-6, April 1964.
11. "The Magnetic Rod - A Cylindrical Thin Film Element," Meier, D. A. and
Kolk, A. J., Large Capacity Techniques for Computing Systems , pp. 195-212,
Macmillan Publishing, New York, 1962.
12. "Investigation of a Woven- Screen Memory System," Davis, J. S. , and Wells,
P. E., Proceedings FJCC, pp. 311-326, Vol. 24, Las Vegas, Nevada,
Nov. 12-14, 1963.
13. "Woven Thin-Film Wire Memories," Maeda, H. and Matsushita, A., 1964
Intermag Conference Proceedings, pp 8-1-1 - 8-1-6, April 1964
2-188
14. "The Plated-Wire Memory Matric," Oshima, S., Futami , K. , and
KamibayashI, T. , 1964 Proceedings Intermag Conference, pp. 5-1-1 -
5-1-6, April 1964.
15. "Laminated Ferrite Memory," Shahbender, R. , Wentworth, C,
Hotchkiss, K. , Li, K. , and Rajchman, J. A., Proceedings, FJCC,
Las Vegas, Nevada, Vol. 24, pp. 77-90, November 12-14, 1963.
16. "An Approach Towards Batch-Fabricated Ferrite Memory Planes,"
Bartkus, E., Brownlow, J., Crape, W. , Elfant, R. , Grebe, K. , and
Gutwin, 0., IBM Journal of Research and Development, pp 17-176, Vol.
8, No. 2, April 1964.
17. "System and Fabrication Techniques for a Solid State Random Access
Mass Memory," Fuller, H., McCormack, T. , and Battarel, C, 1964
Proceedings Intermag Conference, pp. 5-5-1 - 5-5-4, April 1964.
18. "Digital Computer Peripheral Memory," First (Quarterly Report,
(July 1, 1963 - September 30, 1963), U.S.A.E.R. £■ D.L. Contract,
No. DA 36-039-AMC-03248 (E) prepared by RCA Laboratories.
19. "Digital Computer Peripheral Memory," Second Quarterly Report, (Oct, 1,
1963 - Dec. 31, 1963), USAERS-DL Contract, No. DA 36-039-AMC-03248 (E) ,
prepared by RCA Laboratories.
20. "Ultrasonic Approach to Data Storage," Gratian, J. W. and Freytag, R. W. ,
Electronics , Vol. 37, No. 15, pp. 67-72, May 4, 1964, McGraw Hill
Publ i cation.
21. "Review and Survey of Mass Memories," L. C. Hobbs, Proceedings FJCC,
pp. 295-310, Vol. 24, November 12-14, 1963.
22. "Comparison: Major Types of Mass Memories," L. C. Hobbs, Data Systems
Design , Vol. 1, No. 1, pp 18-21, January 1964.
2-189
2.6 COMPONENTS AND PACKAGING
Components and packaging techniques have been investigated and are
discussed together as intimately interrelated topics. The method of
packaging arrays of components cannot be considered independent of the
nature of the component itself. On the other hand, the selection of
specific types of components imposes unique requirements on the packaging
techniques. The ultimate goal is the maximum degree of batch-fabrication
possible to permit relatively large segments of a computer, or other
digital equipment, to be fabricated as a unit in a single set of processing
operations .
Until recently, it has been necessary to package individual discrete
components (e.g., transistors, diodes, resistors, capacitors, etc.) into
circuit arrays by techniques such as printed circuits or welded connections,
and to further interconnect groups of these modules into subassemblies.
Frequently cables and cable connectors are used to interconnect such
subassemblies into units of equipment. With new types of components,
such as integrated circuits, it is no longer necessary to interconnect
physically discrete individual components into a circuit module. Steps
are being taken toward developing techniques for fabricating combinations
of circuit modules without requiring separate packaging and interconnection
operations. It is believed that ultimately major subassemblies will be
made as a single unit by batch-fabrication processes.
One of the largest problems facing the widespread application of
integrated circuits is that of efficient packaging and finding suitable
interconnection techniques. This involves questions such as the maximum
size of a "throw away" unit, spares and logistics, maintenance, and
flexibility. For example, Is It advantageous to have a computer
fabricated with a few hundred modules of packages, each of which Is
unique, to minimize the total amount of equipment and cost? Or, Is
it desirable to use five or ten times as many modules or packages,
of perhaps 20 - 50 different types, to enable stocking of less spares?
These types of questions have to be answered before a 1970 system
Is designed.
2-190
2.6.1 Classification of Components
Components discussed in this section are the logical components
used in mechanizing the control, arithmetic, and other logical functions
in a digital system. Memories and peripheral equipment are not considered
components in the context of this discussion. Components considered
here are those necessary for performing logical operations, providing
temporary storage of the results of logical operations, and amplifying
or shaping signals. Examples of these are diode gates, flip-flops,
and transistor amplifiers respectively.
These components can be classified by whether they are active or
passive, by the method of fabrication, by whether they are electronic,
magnetic or optical, by what circuit or logical function they perform,
and on the basis of other characteristics, such as speed or cost. The
major classifications used in this discussion are: electronic, magnetic,
or optical. Most of the discussion is devoted to electronic components
which are classified on their method of fabrication. One method of
fabrication is by the use of discrete components such as individual
transistors, diodes, resistors, and capacitors. Several other methods
of fabrication are classed under the general category of "integrated
c i rcu i ts" .
Integrated circuits can be divided into four categories, again
based on the method of fabrication:
1) Hybrid circuits in which passive elements are printed on
a ceramic substrate, and discrete (but unpackaged) active
components are connected to printed interconnections on
the same substrate. This combination is then packaged as
a single unit« (The "solid logic" components used in the
new IBM 360 system are examples of this method of fabrication).
2-19
2) Monolithic integrated circuits in which a number of active
elements (e.g. transistors and diodes) and the associated
passive elements to perform a specific circuit function, or
set of circuit functions, are fabricated by diffusion processes
in a single silicon chip.
3) Hybrid monolithic thin-film circuits in which active elements,
and possibly certain passive elements, are diffused into a
silicon chip as in the preceding case but with additional thin
film passive elements and interconnections fabricated on top
of the same silicon chip by vacuum deposition processes.
k) Active thi n-f i Im ci rcui ts in which both the active components
as well as the passive components are fabricated by vacuum
deposition of thin-film elements.
All-magnetic logic components and each of the types of integrated
circuits listed above are discussed in further detail in Section 2.6.5
2.6.2 Sources of Information
2.6.2.1 People and Organizations
Organizations with whom components and packaging techniques have
been discussed include the following:
Motorola Semiconductor Div, Integrated circuit sense amplifiers
Phoenix, Arizona Integrated circuit storage registers
Monolithic integrated circuits
Hybrid integrated circuits
Remington Rand UNIVAC Hybrid integrated circuits
St. Paul, Minn. Integrated circuit reliability S-
fai lure analys i s
Packaging techniques
Control Data Corp. Integrated circuit applications
St. Paul, Minn. Packaging techniques
2-192
Autonet ics
Anaheim. Ca] i f .
Bunker-Ramo Corp.
Canoga Park, Cal if.
RCA Laboratories
Princeton, New Jersey
Hughes Semiconductor Div.
Newport Beach, Calif.
Sy 1 vani a
Waltham, Mass.
Monolithic integrated circuits
Integrated f i eld-effect- trans i stor
ci rcu i ts
Integrated circuit packaging
techn i ques
Monolithic integrated circuits
Integrated circuit sense amplifiers
I ntegrated field -effect- trans i stors
Metal oxide semiconductor integrated
c i rcu i ts
Active thin film integrated circuits
Integrated circuit packaging
techn iques
Monolithic integrated circuits
Hybrid integrated circuits
Active thin film integrated circuit
Tunnel diode circuits
ONR
Washington, D.C.
Optical components
RADC
Rome, New York
Optical components
SR
Menlo Park, Cal if.
Fairchild Semiconductor
Mountai nvi ew, Calif.
Optics Technology
Belmont, Cal i f .
Al 1 magnet ic logic
Cellular logic for integrated
c i rcu i ts
Electron beam fabrication
Fluid log ic
Monolithic integrated circuits
Hybrid integrated circuits
Optical components
NASA
Washington, D.C.
National Security Agency
Fort Mead, Va .
Active thin-film integrated
ci rcu i ts
Monolithic integrated circuits
Packaging techniques
Optical techniques
Integrated circuits
2-193
Information obtained during discussions with personnel of these
organizations provided a basis for much of the information in other
sections of this report. The opinions of experts in specific areas
in these organizations were solicited concerning the advantages,
disadvantages, limitations and future prospects for different circuit
and packaging techniques.
2.6.2.2 Literature
An extensive list of references pertinent to the study of components
and packaging techniques is given in the Bibliography. Many of these
references have been scanned but only a limited number of the more
important ones have been studied in detail to date. A study of these
references has contributed to the material presented in subsequent
parts of this section, and direct quotations have been used where
noted. The more pertinent and important of these references will be
studied in detail during the remainder of this investigation and new
references will be Included to reflect material published or discovered
subsequent to the preparation of this Bibliography.
2.6.3 Components and Packaging Characteristics forANTACCS
It is anticipated that relatively high-speed components with high
reliability will be required for ANTACCS and AMTACCS equipments. These
components and packaging techniques will have to be chosen on the
basis of their ability to meet specifications for shipboard and ground-
based military systems. Characteristics of applicable components and
packaging techniques will be compared and evaluated but the choice
of specific characteristics will depend upon the details of machine
organization and the results of the requirements analysis. Components
and packaging techniques Investigated will Include those necessary
for meeting all requirements both within the central computer and In
peripheral and auxiliary equipment.
2-194
2.6.4 Applications of Components in the Naval Environment
It is anticipated that all components and packaging requirements
for shipboard and ground-based military environments for ANTACCS and
AMTACCS type systems will be covered by this study. Emphasis will
be placed on digital techn iques, but certain appropriate analog
techniques (e.g. memory sense amplifiers, analog-to-digital converters,
digital-to-analog converters, etc.) will be considered. The specific
applications of these components will be considered in further detail
during the remainder of this study.
2.6.5 Current Status Review
In most available commercial and military computers, logical
operations are mechanized with discrete semiconductor circuits - usually
transistor flip-flops, transistor amplifiers, and transistor or diode
gates. Few such discrete semiconductor circuits will be in use in 1970.
Logical components investigated so far during this study include:
Cryogenic logic
Fluid logic
Optical logic
Semiconductor logic using special elements (e.g. tunnel diodes)
Al 1 -magnet i c log i c
Semiconductor integrated circuits
Most of the time to date has been devoted to Investigating semi-
conductor integrated circuit techniques for digital equipment. All of
the types of components listed above are discussed briefly in this
section, but the major part of the discussion is devoted to integrated
circuits. Packaging techniques are also discussed in some detail.
In analyzing components and packaging techniques for a 1970 system,
primary consideration should be given to their adaptability to batch-
fabrication techniques. Microelectronics and batch-fabrication techniques
2-195
are frequently associated and discussed as though they were synonymous
However, there is a distinction in that the term "microelectronics"
places emphasis on miniaturization and small size; whereas the term
"batch-fabrication" places emphasis on methods of fabrication, these
are methodized to permit relatively large numbers of elements to be
fabricated in a "batch" without the necessity for individual handling
of discrete elements.
A number of microelectronic techniques are not adaptable to batch-
fabrication. On the other hand, techniques necessary to achieve batch-
fabrication processes tend to lead to small sizes of individual elements
and hence to microelectronics.
2 .6.5 . 1 Fluid Logi c
Fluid logic is usually mechanized by hydraulic or pneumatic
systems in which small mechanical movements are used to switch the
path or flow of a hydraulic or pneumatic media e.g. oil or air . The
major disadvantages of fluid logic are the relatively slow response
time (milliseconds) and the size and weight compared to equivalent
Integrated semiconductor circuits. Fluid logic offers advantages in
some adverse environments such as high temperature, electromagnetic
fields, and nuclear radiation. Fluid logic may also offer advantages
(e.g. lower cost) in applications where the initial input information
Is in a mechanical form and a mechanical output is required. An
example of this is a desk calculator where the Initial inputs are key
depressions and the outputs are mechanical printing operations. Fluid
logic may be applicable to some shipboard funct ions, such as weapon
direction and ship's control. However, fluid logic will not be
competitive with integrated circuits for logical functions In the
central processor and auxiliary equipment for the Information
processing portion of a 1970 ANTACCS system.
2-196
2. 6. 5.2 Cryogenic Logic
Cryogenic techniques have been discussed previously in the section
of this report dealing with memories. Cryogenic logic and switching
devices, such as the cryotron, have been proposed for computer use
for approximately ten years. During this time, they have not been
proved superior to semiconductor techniques. Although there is some
controversy concerning this, most workers in the field concede that
logical components are the least likely application for cryogenic
2
techniques in a computer . Most of the cryogenic research and develop-
ment work remaining at this time is concentrated on associative and
large-capacity memories. Cryogenic logic and switching is considered
primarily as an adjunct to a large-capacity or an associative cryogenic
memoryc The use of cryotrons for the selection tree in a large-capacity
cryogenic memory is an example. It is very unlikely that cryogenic
logic techniques will be competitive with semiconductor integrated
circuit logic in 1970.
2.6.5»3 Optical Logic
There has been considerable interest in the possibility of using
optical logic devices in computers because of the inherent speed
theoretically possible when working with light. Recent developments
3 if 5
in fiber optics and lasers have accelerated this interest * ' .
Some of the characteristics of lasers that make them attractive for
computer use are:
1) The output is coherent and monochromatic
2) Very high frequencies are possible
3) The beam is highly col 1 imated
4) High-power intensity
5) Capable of either continuous or pulse operation.
2-197
Fiber optics have the capability of conducting light around
curved paths, and hence they offer interconnection possibilities
similar to the use of wires in carrying electrons. However, lasers
and fiber optics will see use in memories and display areas before
they will be successfully used as logical components. Optical logic
techniques offer great promise for the long range future, but they will
not be feasible for use in a 1970 system.
2.6.5'^ Special Semiconductor Elements
A number of unique or special purpose semiconductor devices have
been proposed for use in computers. Of these, the most serious con-
6 7
sideration has been given to the tunnel diode ' . Tunnel diodes have
been proposed for high-speed, smal 1 -capac i ty memories as well as
high-speed logical components. Of the various approaches to "ki lomegacycl e
circuits" tunnel diodes are considered the most practical, although the
rates of approximately 200 - 500 megacycles at which they have been
used do not quite fall in the ki lomegacycl e range. Soon after they
were introduced 4 or 5 years ago, tunnel diodes were considered by
many people to be an exciting solution to the high-speed computer circuit
problem. However, difficulties in working with a two terminal device
such as the tunnel diode have seriously dampened this enthusiasm. The
problems of interconnection techniques for tunnel diode circuits
operating at a frequency of several hundred megacycles have proved to be
difficult. Systems become very expensive as a result of the discrete
mechanical configurations required for interconnections and shielding.
Although tunnel diode logic circuits are feasible, they will not be
competitive with integrated circuit techniques for an ANTACCS type
system in 1970.
2.6.5.5 All-magnetic Logic
Magnetic elements can be used in a digital system for logic as well
as for storage functions. However, magnetic elements have not enjoyed
the widespread use or success as logical elements that they have as
2-198
memory elements. This is, of course, due to the difference in the
nature and requirements of memory components and logic components. A
single word location in a memory is addressed at one time, and a large
driving current, low sense signal, and destructive read out are acceptable.
On the other hand, for a logic element it is necessary to provide some
form of gain, and to sense the state of the device without changing it.
A number of applications for all-magnetic logic and a number of
types of logical configurations and elements have been described in
the literature ' ' ' . All-magnetic logic offers several distinct
advantages including:
1 ) Hi gh rel iabi 1 i ty
2) Radiation resistance
3) High temperature operation
k) Low stand-by power
5) Non volat i le
6) Low power required at low frequencies
7) Cost (in some cases).
The major disadvantages in the use of all -magnetic logic have been
the inherent slow speed and the lack of a steady state output indication.
For most applications all-magnetic logic circuits have not proved them-
selves sufficiently advantageous with respect to either cost or performance
to encourage their use in place of the simpler and more common semi-
conductor circuits.
The characteristics of all-magnetic logic are ideal for certain
applications such as an onboard computer in a deep space probe. In
this application, very low speeds are acceptable and radiation resistance,
low standby power (since the spacecraft is inoperative for the long
cru i se per iods of the mission) and high reliability are important.
However, for the type of applications encountered in an ANTACCS system,
all-magnetic logic is not considered competitive with semiconductor
2-199
circuits - particularly not with integrated circuit techniques for a
1970 system. The radiation resistance characteristic and the high
reliability would be important if future developments prove that
integrated circuits are not as reliable as magnetic logic. However,
it appears that semiconductor integrated circuits will approach the
reliability of all-magnetic logic, and that the cost of these elements
will be less than that for the discrete elements used in magnetic
logic. The speeds of all-magnetic logic are not sufficient for the
central processor. The mu 1 t i -apertured devices found most suitable
for magnetic logic have been limited to rates of a few hundred
kilocycles per second. For other shipboard functions such as
peripheral equipment, weapon direction systems, and ship's control
systems where high speeds are not required, magnetic logic may prove a
good choice. It is possible that new developments in thin-film integrated
magnetic circuits in the next few years may enhance the performance
and significantly reduce the cost of all-magnetic logic.
2.6.5.6 Integrated Circuits
Integrated semiconductor circuits are by far the outstanding
candidate for the logical mechanization of a 1970 ANTACCS system.
Integrated circuits have been proved feasible and successful, and
are currently being used in several military computers. Hybrid
integrated circuits are used in the Remington Rand CP667 computer
12 13
for NTDS and in the new IBM System 36O commercial computer '
Monolithic integrated circuits have been used in the Autonetics
Monica Computer, and are being used by Litton in the computer for
the new F-lll (TFX) . Estimates of the military use of integrated
circuits range from 40 to 50% of all military electronics in 1970
to approximately 75% in 1973 ' * • Since digital circuitry is
more adaptable to integrated circuit techniques, the estimates are
even higher for computer and data processing equipment - approximately
707o by 1970. Actually, this figure will probably be closer to 90%
(not including the memory) for new digital equipment designed to
become operational in 1970.
2-200
Integrated circuits are not basically new components In the sense
that lasers are, but rather they represent radically new methods of
fabricating and packaging semiconductor circuitry. The reduction in
the number of discrete components, resulting from fabricating complete
circuits as a single component, offers significant advantages in terms
of reliability, cost, and size. Batch-fabrication of volume quantities
of integrated circuits will result In significantly lower costs than
is achieved by the present printed circuit and hand wiring of basic
components and circuit modules.
Mr. J. M. Bridges, of the Department of Defense, states that
"a semiconductor integral circuit containing the equivalent of some
20 parts displays the same failure rate as a single conventional
transistor," and he predicts failure rates of approximately 0.0001%
1 8
per 1000 hours . Failure rates as low as 0.001 to 0.0002% per 1000
hours are anticipated for the advanced Minuteman computer, and a number
of estimates place the ultimate reliability of monolithic Integrated
]k 19
circuits as 0.0001% failures per 1000 hours ' . Dr. Noyce of
Fairchild Semiconductor has described the actual reliability experienced
19
on two specific aerospace computers as follows:
"We have data on two operating medium-sized computers that
use integrated circuits. The first Is the Apollo guidance
computer, designed by MIT and built by Raytheon. it has
accumulated 19 million operating hours on its integrated
circuits, in which time two failures have occurred--an initial
failure, and the other a failure, external to the package,
that was caused by moving the computer. The second system,
the MAGIC 1, an airborne computer built by the AC Spark Plug
Computer Division, has accumulated 15 1/^ million hours with
two failures. Fairchlld's in-house life-test program, with
33 million total operating hours, has had a total of eight
failures; of these, five accumulated during the first 6 2/3 million
hours and only three occurred on more recent units during the
last 26,33^,982 hours. These data are not extrapolated from
accelerated tests, but are actual, observed operational failure
rates, and Include early production units In some cases.
Considering the complexity of the function performed by these
circuits, the integrated circuit equipment today is ten times
more reliable than its discrete component counterpart."
2-20
The higher reliability of integrated circuits results from the
fact that there are fewer individual components, circuits are of smaller
size, there are fewer connections of dissimilar metals, most connections
are made by vacuum deposition, and there is less handling of components.
Based on considerations of reliability, cost, size, weight, and
environmental conditions, it is reasonable to expect that integrated
circuits will account for almost all of the logical components in a
1970 shipboard or ground-based military system.
There are four basic types of integrated circuits although these
are sometimes called by different names and in some cases grouped
differently. The term hybrid is particularly confusing since it Is
applied to thin-film passive components with discrete active components
and to thin-film passive components with monolithic active components.
These four types are:
13
1) Hybrid Discrete Thin-film (or Thick-film) Circuits
In this type of circuit, passive elements, such as resistors
and capacitors, are printed on a ceramic or glass substrate
by either vacuum deposition of thin-film elements, or by
printing of thick-film elements In a process similar to
silk screening. Discrete (but unpackaged) active components
are connected to printed or deposited interconnections on
the same substrate. The combination Is then packaged as a
single unit. This is an interim type circuit that was
developed before monolithic and hybrid monolithic circuits
were technically feasible for large scale production.
This type of circuit offers the advantage that the passive
components can be made cheaply with tightly controlled
tolerances. Relatively large values of capacitance can be
fabricated and resistance values can be maintained within
a few percent. As a result, this approach Is more adaptable
to linear circuits, such as differential amplifiers and
2-202
analog circuitry, at present than is the monolithic integrated
circuit. This type of circuit also has the advantage that
there are no interactions and parasitic capacitances between
the different elements as is the case for the monolithic
integrated circuit. It has the disadvantage that the
active elements must be handled as discrete elements. The
reliability is probably not as high, due to the handling of
the active elements and the soldering of these elements to
the printed interconnections on the substrate. The cost will
be higher and large arrays of logical circuits cannot be
batch-fabricated. It is believed that this type of circuit
will phase out before 1970 with preference being given to the
second and third type of integrated circuits discussed below.
20
2) Monolithic Integrated Circuits
This type of circuit is completely integrated. Active elements
(e.g. transistors and diodes) and the associated passive
elements (e.g. resistors and capacitors) necessary to perform
a specific circuit function or set of circuit functions,
are fabricated by a series of diffusion processes in a single
silicon chip. This circuit has the advantage that all
components in the circuit are made during the same series
of operations, and that multiple circuits of this type can
be batch-fabricated in a single set of operations.
This type of circuit should ultimately be cheaper to fabricate
and more reliable due to the ability to make all inter-
connections by vacuum deposition processes. It is more
adaptable to the batch-fabrication of large interconnected
arrays such as a major segment of an arithmetic unit.
There have been three major disadvantages with respect
to monolithic integrated circuits to date:
a). The interaction between semiconductor elements
diffused in the same silicon chip and the resulting
parasitic capacitances.
2-203
b) . Difficulty in maintaining resistor tolerances better than
approximately 20%.
c) . Difficulty in fabricating capac i tanc i es of more than a
few micro-microfarads.
The yield of this type circuit has not been as satisfactory
since any individual bad element makes the entire circuit bad.
It is difficult to get accurate information on the yield
experienced by manufacturers, but estimates range from
approximately 1% to 20% for present high-grade military
type circuits with yields of 50 to 90% predicted for the
future. Monolithic integrated circuits are well suited to
digital applications where component values are not as critical,
but they are not satisfactory for most types of linear circuits
at present because of the interactions and the difficulty in
controlling tight tolerances. Intensive research and develop-
ment efforts are being expended on the problems of monolithic
integrated circuits and rapid progress is being made. Both
Signetics and Motorola have reported success in isolating the
components in a monolithic integrated circuit to reduce the
parasitic capacitance. This should increase the speed of
circuits of this type and permit their application in
certain types of linear circuits. It is anticipated that
this type of circuit will be the major integrated circuit
technique used in digital applications within the next
few years .
2 1
3) Hybrid Monolithic Thin-film Circuits
In this type of circuit active elements, and possibly certain
passive elements, are diffused into a single silicon chip
as in the preceding case. However, additional thin-film
passive elements as well as interconnections are fabricated
on top of the silicon chip by vacuum deposition processes.
2-204
This technique combines the advantages of the first type of
hybrid circuit discussed with the advantages of the completely
monolithic integrated circuit. Tight tolerances on resistors
and capacitors can be maintained and relatively large values
of capacitance fabricated while not handling discrete
components. Batch-fabrication of arrays of elements and
circuits in a single set of processes, and higher reliability
resulting from vacuum deposited interconnections are achieved.
With this type circuit, it is possible to obtain many of the
cost and reliability advantages of the completely monolithic
integrated circuit while fabricating higher quality components.
The fabrication of linear integrated circuits, such as
differential amplifiers and other analog type circuits, is
facilitated. Several hundred thousand ohms of resistance
and several hundred mi cromi crofarads of capacitance can be
obtained on an integrated circuit using this hybrid approach.
Resistor tolerances of better than lO^^and capacity tolerances
of two parts per million can be obtained relatively easily.
Higher resistor tolerances can be achieved by "trimming" the
resistors during the test operation.
This technique will be used along with the completely
monolithic integrated circuit for the next five to eight
years at least. Monolithic integrated circuits will be
used wherever possible, with the hybrid monolithic
thin-film circuit being used to complement and supplement
them where higher tolerance components or larger values of
capacitance are required. Unless the isolation problem in
the monolithic integrated circuit is completely overcome, the
hybrid monolithic thin-film approach will also permit higher
speeds .
2-205
22 23
k) Active Thin-film Element Circuits '
In this type of circuit, both the active components and the
passive components are fabricated by vacuum deposition of
thin-film el ements .
Predictions concerning the date at which active thin-film
elements will become feasible vary widely - from "almost
immediately" to "not less than five years". The longer
estimate is probably the more accurate one with the
possible exception of a related device - the metal-oxide-
semiconductor. A field effect transistor can be fabricated
in this way by depositing germanium or silicon on a passive
substrate, depositing an oxide insulator such as silicon
monoxide, and depositing aluminum plates for connections
and distributed capacitance. Cadmium-sulphide is frequently
used Instead of germanium or silicon.
This type of device offers excellent radiation resistance
characteristics and is quite amenable to batch-fabrication
of large interconnected arrays with minimum interaction.
This circuit Is attractive because of Its simplicity. RCA
reports yields of 90 - 95% compared to approximately 20%
for conventional silicon Integrated circuits.
Another advantage of this type of device Is that It Is
Ideally suited to a complementary symmetry type of circuit
because of Its bipolar nature. One field effect transistor
can essentially act as the load line for another field effect
transistor. As the characteristics of one transistor change
due to external conditions, the characteristics of the other
change also, resulting in a lesser effect of the net change.
Although the metal oxide semiconductor type of field effect
transistor can perhaps be used in a 1970 system, it Is
doubtful that any other types of thln-fllm active elements,
such as thin-film transistors, will be In use until later
during the period between 1970 and 1980.
2-206
In considering integrated circuits for logical components, it is
also necessary to consider the type of logical configuration to be
used. The major types are:
1) Direct coupled transistor logic (DCTL)
2) Diode transistor logic (DTL)
3) Resistor transistor logic (RTL)
k) Resistor capacitor transistor logic (RCTL)
5) Transistor coupled transistor logic (TTL)
6) Emitter coupled transistor logic (ECTL) also referred to as
current mode logic (CML, or MECL)-
The choice between these different types of logical circuit depends
upon the function for which the circuit is chosen and the method of
fabrication of the integrated circuit itself. The relative importance
of speed, cost, power, size, and reliability will vary with different
applications and different circuit fabrication techniques. The major
advantages and disadvantages of each type are shown in the Table 2.9
. , 24,25,26,27
below: ' ' '
Logic Circuits
DCTL
RTL
RCTL
TTL
TABLE 2.9
Advantages
Low power
Simpl i ci ty
S impl i c i ty
Better load
Good load
d i str i but ion
Good noise re-
ject ion
High fan out
Low power
High speed
Simpl ici ty
Low power
Di sadvantages
Poor load distribution
Noise sensitive
Low fan-out
Noise sensitivity
Slower speed than DCTL
Slower speed
More complex circuit
Low fan-out
Poor noise sensitivity
2-207
Logic Circuits Advantages Disadvantages
DTL Good noise immunity Two power supplies required,
Good isolation Slower speed,
Good fan- in Low fan-out
capabi 1 i ty
Low power
ECTL Simplicity More critical circuit
Good load parameter
distribution More components
High speed Two power supplies
operation Noise sensitivity
Monolithic integrated circuit application for linear circuits
have not progressed as far due to the problems with interaction between
components, parasitic capacitance, and the difficulty of fabricating
larger values of capacitance. As a result, most of the success in
integrating linear circuits has been with hybrid type integrated
circuits. Differential amplifiers and other types of analog computer
circuits have been difficult to mechanize with monolithic circuits for
these reasons. A good deal of effort has been expended on certain
types of linear circuits for computers--part icularly sense amplifiers
28
for memories
It is believed that satisfactory memory sense amplifiers in monolithic
form will be available within 1 - 2 years. This will have a significant
effect on memory costs for large capacity memories as discussed in the
memory section of this report. Other types of circuits, such as magnetic
memory drive circuitry, have been difficult to mechanize in monolithic
form because of the power handling requirements. The solution to this
problem is not as close as the solution to the sense amplifier problem.
Ultimately, the successful widespread use of integrated circuits
in computers and information processing systems will depend upon the
industry's ability to find new and more effective ways of utilizing
larger arrays of individual circuits. Although significant improvements
can be achieved by replacing the discrete semiconductor circuits with
integrated circuits in present types of logical configurations and
machine organization, new approaches will be required to realize the
ultimate potential of integrated circuits. It will be necessary
2-208
to fabricate groups of circuits in "functional electronic blocks" or in
29 30
some kind of generalized "cellular logic" array ' . The use of larger
function electronic blocks depends upon techniques for making major
segments of a machine more repetitive, so that a relatively large
number of similar blocks can be used. This is possible now in some
arithmetic parts of a parallel machine where successive stages of
registers and adder circuitry are repetitive. However, it is very
difficult at this time in the control parts of a machine where there
is little tendency for repet i t i veness . In a cellular array, a large
number of similar circuits would be fabricated on a chip with appropriate
means of semi -standard interconnections between them. Methods of
designing computers with this type of structure need further investigation.
In either case, problems of redundancy, the ability to work with a
limited number of bad elements, and interconnection techniques need
extensive work.
2.6.5«7 Packaging
It has been pointed out that one reason for the increased
reliability of integrated circuits Is that groups of elements can
be interconnected by vacuum deposition processes rather than by
27 31 32
soldering, welding, or crimping ' ' . The use of vacuum deposition
techniques can lead to the formation of molecular junctions at the
points of interconnections rather than the Interfaces that result from
other methods. The vacuum deposition of interconnections also removes
much of the human element. This advantage has been described by
Mr. McKenzle of Electronics magazine as follows:
"Whereas welding or soldering constitute a weakening of reliability,
owing to possible carelessness or ineptitute of a technician, the thin-film
applied through a fixed mask would necessarily provide automatic and
uniform interconnection.
Present Interconnection practice involves many methods of making
joints and the connecting leads themselves are of materials chosen as
best suited for joining. Hand soldering may always be used for a
number of larger joints or touch-up work, but as the size of units
decreases the uncertainty as well as the damage sometimes caused will
continue to curtail use of hand soldering.
2-209
Automatic dip soldering and flow soldering involve certain hazards
such as overheating, corrosion from flux, and particles of excess solder
The joints are good only to the melting point of the solder used.
Special techniques such as the use of solder preforms and hot air, are
continually under investigation but the limitations of the soft-solder
joints are understood and efforts are directed to better methods of
joi ni ng .
Welded circuits can be successfully made and the joints hold up
to temperatures of about 1,500 F. Initial problems of obtaining
satisfactory welds with tinned copper, brass and nickel-iron alloy
wires have been largely eliminated through the use of nickel, nickel-
clad copper and stainless clad copper. Improvements in welding
techniques have produced successful joints even with formerly difficult
materials. Data are still lacking on the definite improvement in
reliability of the welded over the soldered joint but it may be as
high as 20 to 1 ." 31
A large percentage of the bulk of present day computers is
composed of interconnections, connectors, and cables. The use of
functional electronic blocks or cellular logic permitting the batch-
fabrication of interconnections for large groups of circuits will
greatly alleviate this prob lem^^' ^°'^^'^^' ^^ .
The inter-connection of integrated circuits is another possible
application for lasers. Many of the problems of soldering and welding
inter-connections can possibly be overcome by using a laser micro-
welding technique . The use of a laser for welding does not require
high vacuum equipment as does electron beam welding equipment, no
foreign materials are introduced into the joint as in soldering, and
heating of the elements is not necessary. No pressure is applied to
the joint, and the laser beam can reach places that are inaccessible to
other welding techniques. The use of the high energy beam from the
laser for welding purposes has been demonstrated and is being further
i nvest i gated .
Interconnection between integrated circuit blocks has been
accomplished by a number of techniques including the use of multi-
layer boards, a cord wood structure, and micromodule techniques.
2-210
The multi-layer board approach seems to be the most widely acceptable
at this time . The choice of interconnection technique also involves
a number of questions other than the actual making of i nterconnect ions--
What is the minimum size of throw-away package? What is the effect
on maintenance and spares? How is layout and organization of the
machine affected? How adaptable is the technique to batch-fabrication?
Is adequate heat transfer provided? What is the volumetric efficiency?
Another problem in the assembly of groups of integrated circuits,
as in any other type of electronics, is that of cooling. One interesting
approach to this problem is to provide a completely controlled atmosphere
by immersing all the components in a liquid such as Freon. The Freon
can be maintained at a constant temperature by external water cooling to
permit close control of the temperature around the individual components.
It also keeps foreign substances such as dust and humidity away from
the components. The ability to control the temperature and environment
in which the components are working simplifies this basic circuit design
and permits higher performance circuitry by removing the necessity for
working over a large temperature range.
The problems of packaging and interconnection of basic circuit
modules will be investigated further in the remainder of this study.
2.6.6 Availability of Components in the 1970-1980 Period
Completely integrated circuit components capable of fulfilling the
requirements for the central processor and peripheral equipment will be
available prior to 1970. All magnetic logic for slower speed application
will be available but may not be competitive. It is unlikely that
optical logical components will be available until at least the mid
1970's. Cryogenic logic and special semi-conductor devices such as
tunnel diodes will not be competitive with integrated circuits for
ANTACCS type applications.
2-2
Although there is much work to be done in the area of packaging
techniques and integrated circuit components, it is believed that
adequate techniques will be available for use in a 1970 system.
It is possible that two basic types of logical components will
be used in an ANTACCS type system in the 1970 time period. One of
these will be high-speed semiconductor integrated circuit components
with large fan-in and fan-out capabilities for mechanizing the central
processor and other high-speed parts of the system. For economy purposes,
a second type of circuit might be used in peripheral equipment and slow-
speed applications. These circuits may be either a slow-speed semi-
conductor integrated circuit type component or perhaps all -magnetic
logic components. If the cost differential is not significant between
the two categories of components, the high-speed integrated circuit
components may be used even in the slow-speed peripheral equipment to
provide a higher degree of standardization and to reduce the spares and
maintenance requirements.
2.6.7 Limitations of Present or Planned Components and Packaging Techniques
Limitations of cyrogenic logic, fluid logic, and special semi-
conductor elements have been discussed previously. The primary
limitations of all-magnetic logic is one of speed, and it is possible
that this type of component will be used in slow-speed applications.
No limitations on the availability or capability of semi-conductor
integrated circuit components for applications in ANTACCS type equipment
are foreseen for the 1970 period. Integrated circuits will be capable
of meeting and exceeding all the requirements for digital type circuits
with the possible exception of high powered output components.
2-212
2.6.8 Recommended Developments to Meet ANTACCS Needs
No additional developments in the area of logic components are
needed to meet ANTACCS needs for the 1970 period. Adequate research
and development efforts are currently underway on integrated circuits
to assure the necessary components for a future NTDS system. However,
additional effort is needed to develop improved packaging techniques
and packing philosophy for the optimum utilization of integrated circuit
techniques and batch-fabrication processes. This will require work
not only in the specific area of packaging techniques, but also in the
areas of machine organization to permit types of logical configurations
that are readily adaptable to the batch-fabrication of large arrays of
c i rcu i ts .
2.6.9 Evaluation Criteria Recommended
Recommended criteria for evaluating components and packaging
techniques will include the following:
Type of logic
Type of circuit elements
Type of fabrication
Number of active elements per circuit package
Number of passive elements per circuit package
Approximate cost per circuit package
Propagation delay
Power dissipation
Power requirements
Permissible levels of logic
Fan-in and fan-out ratios
Noi se sens i t i v i ty
Nature of active elements
Stand-by power requirements
Operating power requirements
Susceptibility to nuclear radiation effects
Susceptibility to electromagnetic interference
Generation of electromagnetic interference
Susceptibility to shock and vibration
Susceptibility to humidity
Operating temperature range
Special requirements (e.g. cooling or refrigeration)
Approximate date of first production quantity applications
Batch-fabrication techniques
2-213
Some of these will rule out certain types of components without
the necessity for detailed comparisons. Applicable components will be
compared and evaluated on the basis of those characteristics that
directly affect the relative value or importance of competitive components.
For example, components will not be compared on the basis of their suscepti-
bility to nuclear radiation effects, but this will be cited as an advantage
of specific techniques where applicable. On the other hand, the propa-
gation delay, or fan-in and fan-out ratios, will probably be compared
in detail for different types of components or circuit configurations.
2.6.10 Conclusions and Recommendations
Most of the digital parts and a large percentage of the analog
parts of an ANTACCS system for 1970 will be mechanized with semi-
conductor integrated circuits. Emphasis should be placed on batch-
fabrication techniques, not only with respect to the circuits themselves,
but with respect to machine organization approaches that permit the
fabrication of large arrays of circuits in a single set of processing
operations. Further consideration will be given to specific uses of
hydraulic logic and all-magnetic logic during the remainder of this
study, but integrated circuits are considered to be the primary
candidate for mechanization of the 1970 system.
2-214
References; Circuits & Packaging, Section 2.5
1 "Pneumatic Log" l-IV, E. L. Holbrook, Control Engineering,
July, August, November 1961, and February 1962
2 "The Case for Cryogenics?", W. V. Ittner, Proceedings 1962 FJCC,
pp 229-231, Philadelphia, Pa., December 1962
3 "Fiber Optics and the Laser", N. S. Kapany, paper presented at
the New York Academy of Sciences Conference on the Laser,
New York, New York, May 4-5, 1964
4 "The Status of Optical Logic Elements for Nanosecond Computer
Systems," J. T. Tippett, 1963 Pacific Computer Conference, IEEE,
Pasadena, Calif., pp 47-53, March 15-16, 1963
5 "Possible Uses of Lasers in Optical Logic Functions, "C . Koster,
1963 Pacific Computer Conference, IEEE, Pasadena, California,
pp 54-62, March 15-16, I963
6 "A Survey of Tunnel-Diode Digital Techniques," R. C. Sims,
E. R. Beck, Jr., and V. C. Kamm, Proceedings of the IRE, Vol 49,
No. 1, pp 136-146, January I96I
7 "300 mcs Tunnel Diode Logic Circuits," M. Cooperman, I963 Pacific
Computer Conference IEEE, Pasadena, Calif., pp I66-I86, March
15-16, 1963
8 "Design of an All Magnetic Computing System," H. D. Crane and
E. K. Van DeRiet, IRE Transactions on Electronic Computers,
Vol EC-10, No. 2, pp 207-232, June I96I
9 "The Case for Magnetic Logic," J. Rogers, and J. Kings, Electronics,
Vol. 37, No. 17, pp 40-47, June 1, 1964
10 "All Magnetic Digital Circuit Fundamentals," E. E. Newhall, Digest
of 1964 International Solid State Circuits Conference, pp I6-I7,
Philadelphia, Pa., February 1964
11 "All Magnetic Digital Circuits and Application Problems," T. Baker
and C. Dillon, Digest of 1964 International Solid State Circuits
Conference, pp 18-19, Philadelphia, Pa., February 1964
12 "Big Computer Goes in Small Package," Electronics, pp 28-29,
March l4, 1964
2-215
13 "Solid Logic Technology: Versatile, H i gh-Performance Microelectronics,
E. M. Davis, W. E. Harding, R. S. Swartz, J. J. Korning, IBM Journal
of Research & Development, Vol 8, No. 2, pp 102-114
]k "Digital Computer Aspects of Integrated Circuit Applications,"
R. C. Platzek and H. C. Goodman, Proceedings Nat ' 1 Winter Convention
on Military Electronics, Los Angeles, Vol III, pp 2-3^ - 2-53,
February 5-7, 1964
15 "Microelectronics - Where, Why, and When," E. P. O'Connell and
J. S. Brauer, Proceedings Nat ' 1 Winter Convention on Military
Electronics, Los Angeles, Calif., Vol Ml, pp 2-1, Feb 5-7-19/4
16 "1964: The Year Micro Circuits Grew Up," Electronics, pp 10-11,
March l4, 1964
17 "The Economic Impact of Integrated Circuitry," P. E. Haggarty,
IEEE Spectrum, Vol 1, No. 6, pp 80-82, June 1964
18 "Government Needs and Policies in the Age of Microelectronics,"
J. M. Bridges, The Impact of Microelectronics , pp 31-40,
McGraw Hill Publishing Co., New York, New York, 1963
19 "Integrated Circuits in Military Equipment," R. N. Noyce, IEEE
Spectrum, Vol 1, No. 6, pp 71-72, June, 1964
20 "Monolithic Integrated Circuits," Ac B. Philips, IEEE Spectrum,
Vol 1, No. 6, pp A-3 - 101, June 1964
21 "Integrated Linear Circuits," D. Bailey, Electronic Products,
pp 50, June 1964
22 "The Future of Thin-Film Active Devices, Charles Feldman,
Electronics, Vol 37, No. 4, pp 23-26, January 24, 1964
23 "Thin-Film Circuit Technology: Part Ill-Active Thin-Film
Devices," A. B. Fowler, IEEE Spectrum, Vol 1, No. 6,
pp 102-1 1 1, June 1964
24 "Choosing Logic for Microelectronics," A. E. Skoures, ELECTRONICS,
Vol 36, No. 47, pp 23-26, October 4, 1963
25 "Trends in Logic Circuit Design," A. Lambert, Electronics,
pp 38-45, December 6, 1963
26 "Choice of Logic Forms for Integrated Circuits," M. Phelps, Jr.,
Electrical Design News, Cahners Publishing Co., January 1964
2-216
27 Mildata Study, Quarterly Progress Report //I, August 12, 1963 to
November 8, 1963, DA-36, 039-AMC-03275 (E), Honeywell Electronic
Data Processing, 3 December 1962
28 "Utilization of New Techniques and Devices in Integrated Circuits,"
Second Quarterly Report, AF Contract No. AF 33 (657)-11l85
(Pacific Semiconductor Inc.) 1 August 1963 - 31 August 1963
29 "Interconnection and Organization of Functional Electronic
Blocks," H. Winsker and R. Maclntyre, l6th Annual
National Aerospace Electronics Conference, May 1964
30 "Cellular Linear - Input Logic", R. C. Minnick and R. A. Short,
Final Report on AFl 9(628) -448, Project 4641, Tank 4641C1, Stanford
Research Institute, Feb., 1964
31 "Modern Electronics Packaging, A. A. McKenzie, Electronics,
PP 33-48, February 7, 1964
32 "Failure Modes in Integrated and Partially Integrated Micro-
electric Circuits," G. P. Anderson and R. A. Erickson,
Proceedings of Second Annual Symposium on the Physics of Failure
in Electronics, Sept. 25-26, 1963
33 "Flip Chips Easier to Connect," E. Q. Carr, Electronics,
pp 82-84, October 18, 1963
34 "Interconnection of Integrated Circuit Flat Packs In Autonetics
Improved Minuteman Program," E. F. Harman, Autonetics, Pub No.
T4-358/33
35 MICRO Electronics , E. Keonjian, McGraw Hill Publishing Co.,
New York, New York, 1963
36 "Laser Welding for Microelectronic Interconnections," H. Rischall,
Jo Shackleton, 1964 Electronic Components Conference, Washington, DC
2-217
2.7 ADVANCED USAGE TECHNIQUES
For purposes of this present definition, advanced usage techniques
are construed to be those which, while they may be beyond the "experimental"
stage, have as yet no wide application of a pragmatic sort. Certain of
the techniques are hardware-oriented. That Is, the basis for a new
family of usages or for a programming philosophy or problem approach
may arise because of the availability of new hardware features or
departures from customary or ordinary logic design. The inclusion of
interrupt logic In general purpose computers, for Instance, made the
whole field of real-time and on-line applications workable and practical.
Similarly the stored-logic design of the computer of the BRN-3 navigation
set made the use of Interpretive programming practical In that appli-
cation. Such programming had long been judged "impractical" on the
conventional digital computer.
2.7.1 Classification of Advanced Usage Techniques
Based on preliminary investigations and searches of the literature,
the following categories of investigation have been established:
2.7.1.1 Heuristics and Machine Learning
The apparent applicable advances in heuristics programming and
problem formulation, as related to command and control seem to
indicate that little of a generally pertinent nature is to be found
In this area. The same is true of machine learning. Therefore, these
two areas have been combined.
2.7.1.2 System Diagnosis
The rapidly-growing importance of this topic, particularly In
complex data systems, which include one or more digital computing
modules, warrants its conclusion among the advanced usage techniques,
and emphasis on its study.
2-2
System self-diagnosis is, in some measure, hardware-determined.
The abilities for accomplishing diagnosis under computing module
control are provided, usually, by the existence of the proper kinds
of communication paths, and an interrupt structure or its equivalent
which permits the proper level of query and response within the
system.
But a good deal of diagnosis is found in properly-designed pro-
grammi ng .
Certain kinds of diagnosis are not fault-oriented. There may be
a diagnosis in real time, as in the Quotron stock-quoting system which
analyzes system traffic load and permits deferring of low priority or
1 ess- important messagesc
Among the topics of interest in diagnosis are the following:
1) Preserving memory contents during power failure or other
catastrophic failure.
2) Cycling through a pre-determl ned set of tests, either in
free time or in real time.
3) Ability to recycle tests arbitrarily under operator direction
k) The use of redundancy as a diagnosis tool, and to permit
system graceful degradation.
5) The use of back-up systems of equal or less abilityc
6) The use and design of background diagnostic programs in
real-time systems.
All of the above apply not only to on-line diagnosis of the
computational and control sub-system, but also to on-line diagnosis
of communications, sensor and weapons subsystems.
2-219
2.7.1.3 Pattern Recognition
Pattern recognition work in digital computers is progressing
rapidly enough to warrant its inclusion here and the attendant
expenditure of time. Pattern recognition is a broader subject than
character recognition. Character recognition may be thought of as
an important subset of pattern recognition in which the set of patterns
(font) is known and bounded. ., Neither of these is necessarily so in
pattern recognition, since one task conceivable may be the determination
of similarities in two or more patterns, none of which is previously
known .
Among the techniques now used in experimental pattern recognition
are the construction of Boolean matrices in which the pattern is, in
essence, described as an array of ones and zeroes (blacks and whites).
Various operations, such as ANDing and shifting matrices, permit pattern
comparisons and alterations.
Feature extraction also is used to define categorized lines, curves,
and intersections in a pattern, and to assign them to differently-chosen
envelopes or regions which may aid in recognition. Contour analysis,
similar to that performed in cartography, is also used as a technique
in machine pattern recognition.
2.7.1.^ Associative Memories and Related Techniques
The development of associative memories of content-addressable and
other types, together with their obvious applicability to many problems
in command and control, makes emphasis on this area desirable in the
present work. Working modules of content-addressable memories have
been built at various places, notably by Goodyear. Research work has
been done and prototypes have been tested at such places as the
Bunker-Ramo Corporation.
2-220
Content-addressable memories, in wiilch information is stored as
adjoined words, or in whicin search or addressing is possible on the
basis of partial words or information keys, is the usual type of
design,. However, most structures today preclude the finding of dual
entries with certainty. Other possible structures have also beein
investigated, including, for example, those In which indexing, or
relative address structure, becomes a built-in part of the memory
module.
Current designs do not compete in capac ity/cost ratio with con-
ventionally-organized memories, though probably this situation will
change rapidly upward with Increased use of associative memories.
2.7.1.5 Adaptive Systems
Development of these systems, particularly where computers and
display modules work In a single system, makes Inclusion of them
reasonable here.
Adaptive systems are those that change their basis of action with
environment or history. For example, a speed-sensing system element
might have Its calibration changed in real time as a result of
successive observed positions. Adaptive systems have manifest appli-
cations In command and control, and will be investigated In more
detail during the balance of the study.
2«7.2 Current Status of the Advanced Usage Subtask
Literature and Source Compilation is under way on this subtask.
Manning of the subtask has been structured to schedule completion
on or before the end of month ten.
2-22
2.8 COMPUTER SYSTEM ORGANIZATION
Computer system organization deals with the design of the larger
components of a computer system and their relationship to each other
with respect to capability, communication and synchronization. It is an
extremely important subject in computer technology since it is the near-
unanimous opinion of experts in the field that greater strides will be
made in computer efficiency during the next 10 years through organization
than through straight hardware improvements at the component level.
Computer organization Is extremely important for consideration in ANTACCS
because of the stringent requirements that are placed on the system and
because of the inherent complexities of these large-scale, real-time
systems which will implement ANTACCS.
One of the central objectives of the work of computer system
organiz tion is to arrive at a series of recommendations on the
characteristics of a possible future NTDS family of computers. This
objective is motivated as follows:
Sooner or later there will need to be an upgraded family of
computers for future NTDS. Within the next three or four years a
basic decision will likely be made as to whether computers in the 1970's
for NTDS will be upgraded, programmed compatible versions of the present
systems, or a new family of computers with different modular components
and different instruction repertoires will be designed to take their
place. To make that decision it is instructive to examine carefully
the characteristics of a future NTDS computer family should it be
desirable to develop one.
The work on machine system organization is approximately 25%
complete. This part of this Midway Report will consist mostly of an
outline of the subject to be considered and the organization and
intent of the technical efforts. Only one part of the computer
system organization effort is relatively complete. This Is the
2-222
section on "stored logic" or "microprogrammed" computers. However,
in each section of this report on computer system organization there
is a brief discussion of the subject matter to be treated in the
future work.
2.8.1 Classification of Coniou:er Syste:;: Organization
For the purposes of the work thus far, this technical area has
been divided into the following:
Mu 1 t i compuiers and modular concepts
Memory and memory oriented computers
internal organization
New computer trends
Analog/digital hybrids
Existing NTDS computers.
All of these are perhaps self-explanatory, except "new computer
trends". In this area stored logic and microprogrammed computers will
be covered, as well as so-called highly-parallel computers, such as
those of the Solomion type.
!n addition there will be a discussion of: requirements and
applications of computer systems in ANTACCS, recommendations and
comments on a possible future NTDS computer family, and recommendations
and com.m;ents on other types of computer equipments such as specialized
m.emories and peripheral or buffering equipmients, and majority logic,
2.8.2 Sources of Information
2.8.2.1 People and Companies
At this point there has not been a comprehensive survey of people
and companies with respect to this subject. Because of the great
familiarity of the project team with current efforts in the country,
there will not be a great deal of time spent on a thorough survey.
However, it is intended that information and opinions be solicited from
2-223
a number of people close to NTDS computer development in the Navy and
in industry. As a start in this connection, an extensive interview
was held with Mr. Donald Ream of BuShips late in June. Similar interviews
will take place in the next few months.
2.8.2.2 Literature
Since the section on stored logic and microprogrammed computers
is the only section relatively complete at this time, the bibliography
is limited to that technical areao Cited references are listed at the
end of the section. General references are in the bibliography.
2.8.3 liulti Computers and Modularity
Since the advantages of mu 1 t i computers and modularity for ANTACCS
are obvious with respect to reliability and expansibility requirements,
this technical area of computer organization is considered to be very
important. Although this work is now well under way, for the purposes
of the Midway Report the discussion here will be limited to the
organization and topics to be considered.
First of all, definitions and motivations will be presented; that
is, what mu 1 t i computers are and why are they important. Next, existing
hardware configurations will be examined. This will range from the
RW-400 computers and multi-l604 systems for CINCPAC to the D825 systems
for NRL. The manner in which these computers are being used will then
be discussed. Operational factors of mu 1 t i computers will be analyzed
especially as they relate to the ANTACCS environment. Programming
considerations of these larger systems are important since they represent
a new challenge to the techniques. Finally, the future uses in ANTACCS
will be developed .
2.8.4 Memory and Memory-Oriented Computers
Since the memory is the principal part of a computer, it is
important to look at the implications of computer system organization
from the point of view of the memory involved and how it is used.
2-224
In the other parts of the technology work, memories have been discussed
from the standpoint of the hardware configurations and their capabilities
as a component. In this section the computer system is discussed from
the standpoint of the type of memory and how it fits with the total
computer system.
The following topics constitute this section:
Memory types and uses
Memory hierarchies
Memory addressing
Content-addressed or associative memories
Read-only memories
Memory oriented computers
2.8.5 Internal Organization
There are some interesting and important aspects of internal
organization which should be examined. Some of these internal organi-
zation factors have a big impact on the computer's efficiency and the
total computer organization. Some of the topics to be discussed are:
Registers and i nt ra-machi ne communication
Instruction repertoire
I nput/output .
2.8.6 New Computer Trends
2.8.6.1 Stored Logic and- Mi cro programmed Computers
2.8.6.1.1 General
For a number of years the term stored logic has been equated
through usage with microprogramming. Although the literal definitions
of these terms, if they could be agreed upon, might indicate that a
distinction should be made between them, it would be a minor one;
perhaps, simply, a matter of the point of view.
2-225
1 2 3
Historically, the term microprogramming is attributed to Wilkes ' '
The greatest area of agreement concerning the definition among writers
on the subject is that it is difficult. For the most part, the question
is side stepped and earlier definitions cited (as here).
The difficulty is that the concept was initially seen as a radical
departure from conventional design, was implemented somewhat incompletely
(in terms of the original concept) and has since come to be thought of
in terms of these implementations. It, therefore, takes on varying
meanings and associations in different applications of the design.
An ambitious definition is given by Glantz:
"Microprogram (noun)--a program of analytic instructions which
the programmer intends to construct from the basic subcommands of a
digital computer; a sequence of pseudo-commands which will be trans-
lated by hardware into machine sub-commands; a means of building various
analytic instructions as needed from the subcommand structure of a
computer; a plan for obtaining maximum utilization of the abilities
of a digital computer by efficient use of the subcommands of the machine.'
"Microprogramming (verb)--to plan an analytical process in a
pseudo-code which is to be reduced to the subcommands of a digital
computer; to plan an analytical operation in terms of the subcommand
structure of a digital computer; to plan an analysis which will
utilize the subcommands of a computer in an optimum fashion."
The definition hinges on the words "subcommand" and "subcommand
structure" by which is meant simply the manipulation of smaller
elements of logic than is usual. The term pseudo-command, although a
hackneyed term, may mean almost anything, and is used here to indicate
that the code Is different or at least unconventional.
Initially, Wilkes envisioned a design somewhat more specific. He
conceived the possibility of dynamically alterable instruction sets
incorporating the use of two control matrices, a "connection" matrix
and a "sequencing" matrix. One matrix would determine a number of
2-226
control states and the other would select the specific micro-operations
for a particular state. The micro-operations performed would vary then,
depending on which set of control logic was in effect.
From this idea of variable logical machines which depend on the
state of a control matrix, grew the notion that the programmer could
determine a unique order code by combining basic building blocks of
logic (variously called, micro-operations, mi croeommands or subcommands)
The concept Is derived from the fact that the typical machine in-
struction consists of a sequence of basic elementary operations which
are, however, fixed (or wired in), i.e. implemented by hardware.
These sequences are often complicated and Intricate. It was felt that
a basic defect of the conventional machine was the probability of the
superfluous performance of certain of these subcommands serving no
useful purpose in the computation involved.
It was, therefore proposed that the basic machine operations be
made available to the programmer. It was recognized that the selection
of these basic elements would be of paramount importance since the
combinative properties of those chosen would allow the programmer
to develop a powerful logical machine. In effect, the logical design
of the Instruction set would be done by the programmer.
4
Mercer defines microprogramming as "the technique of designing
the control circuits of an electronic digital computer to formally
Interpret and execute a given set of machine operations, as an
equivalent set of sequences of micro-operations, elementary operations
that can be executed In one pulse time."
This would tend to place the responsibility in the hands of the
logic designer and there Is, perhaps, a continuing validity in this
viewpoint. However, the original fascination of the concept lay In
the possibility that the order code could ultimately be chosen at
will by the programmer.
2-227
The "one-pulse" criterion, however, is considerably diluted in
later developments, although, one of the characteristics of stored
logic is the relatively small number of clock pulses per computer
instruction. Furthermore, the concept of programming using individual
microcommands is not perhaps strictly realized (in the sense of a
one-for-one sequencial specification by the programmer). Rather, the
programmer ordinarily specifies that a particular set of micro-commands
occur (perhaps a dozen or so). He may specify explicitly that a
particular one will operate, but usually in combination with others.
He may modify in a sensitive manner his choice of micro-commands, and
may combine them in many ways. In this respect, however, the stored
logic computer may not be so different from the conventional computer
which also may have a sensitive control of operation (with various
modifiers in its instruction word). Indeed, an occasional debunker's
pastime is the "explaining away" of the difference attributed to
stored logic computers in conventional terms. The difference may
turn out to be one of degree.
Nevertheless, we shall attempt to characterize the development
by describing the successful commercial adaptations of the principle
and to indicate certain directions that the development of this concept
may take. For although there does not appear to be precise agreement
as to what constitutes microprogramming or stored logic, and further,
whether intrinsically it is a good design, the effects of the
development to date are undeniable and the future implications are
far-reach i ng.
It will be seen that certain of the early motiviations for this
type of design are no longer so compelling due to other developments
(mostly hardware), and that certain other trends have perhaps re-
inforced the reason for its continued use.
2-228
2.8.6.1 c2 Descriptions of Current Stored Logic Computers
Rather than attempt a rigorous or composite definition of stored
logic, it is perhaps more instructive to consider the common character-
istics of the various stored logic implementations, and to indicate
those attributes that, it is generally agreed, characterize its
development. For, it is more useful to describe it in terms of what
it appears to be now and to derive if possible, what it might presently
be, than to define it in terms of what it was originally conceived to
be. Parenthetically, it might be noted that the earlier thinking in
some respects is the more sophisticated and is perhaps deserving of
attention as a sub-topic in the somewhat neglected field of basic
computer organization.
The subject of stored logic was presented in a series of articles
in the February ]S6k issue of Datamation ' ' ' ' , and the material
contained there was drawn on in preparing this report. The approach
taken in these articles was that of describing the commercial machines
which were currently marketed as stored logic machines; and the concept
is described in terms of these machines. These computers are the
TRW-130/133/530 computers, the PB-440 and the C-8401 .
However, these computers in some respects are as different from
each other as they are from the conventional computer (with which stored
logic computers are invariably contrasted). And, perhaps, even more,
they depart from the original concepts of stored logic and micro-
programming as described by Wilkes. We will, however, examine the
characteristics of these machines briefly, noting the common attributes
and the distinctive features of each.
1) TRW-130(AN/UYK-1), TRW-330, TRW- 133
The TRW-I30 is the forerunner of this family of computers.
It was initially designed under a Navy contract to serve as
a militarized multi-general purpose computer to be used
2-229
primarily for shipboard use. The considerable success of
this computer was probably due more to other design
characteristics (small size, militarized construction,
ruggedness, ability to operate with high reliability under
adverse environmental cond i t ions) , --than to its stored
logic design. It is claimed, however, that the stored
logic method permitted a simplicity of hardware which
would have been impossible to implement economically if
a more standard design had been adopted.
The TRW-130 contains 8192 words of 15 bit storage with a
six microsecond read-write cycle. The TRW-530 is very
similarly organized but has an 18 bit word and certain
additional logical options due to the longer instruction
word. The TRW-133 incorporates the same design as the
TRW-I30 but is three times as fast with a two microsecond
cycle.
Operation may be thought of as occurring on three levels in
the TRW machines; the microcommand level, the machine
instruction level and the interpretive level. Mi crocommands
are not accessible individually to the programmer although
he specifies them in combinations (explicitly and implicitly)
at the machine code level. The machine code command is given
the name Logand ( log ical Comm and ) and occupies one word of
computer memory « A string of logands may be combined to form
a routine called a Logram ( Log i cal Pro gram ) . These routines
which are written in a closed subroutine form operate in a
sequential fashion and are called into operation by the
programmer specifying a list of routines to be operated
(a Logram Calling Sequence). When the computer is used
in this manner it is said to be operating in the interpretive
mode.
2-230
The logical organization features accessible working registers
which are available to the programmer- for the various machine
functions to be performed, sometimes interchangeably. It is
the individual transfers between these registers which are
recognized as the mi crocommands which are defined for this
machine. These registers are used for arithmetic, memory
addressing, logical, and control transfer purposes, and for
input/output and temporary storage. For example, the P
register is used as an addressing register, as an extension
of the arithmetic register, as a shifting register, contains
the quotient in division, the least significant part of the
product in multiplication, and also acts as an instruction
counter for the interpretive level of programming. Incrementing
logic is available and, therefore, indexing and program counter
sequencing may be assumed by these working registers.
The instruction word format (logand format) features an
address option field that provides unusual addressing
flexibility. The address for the current operand is
ordinarily found in one of four of the working registers
mentioned earlier and is specified by the address option
field. Indirect addressing is also available and the
combination properties of this addressing scheme are
designed to minimize addressing overhead.
- The programmer of the lower (machine code) level of coding is given
the name logrammer, presumably because he is composing lograms. He
programs (or lograms) using logands. The term logander, however, is
not valid. The coder who uses logram calling sequences is called a
programmer .
2-23
The instruction word has various formats and may contain two
functional commands (operations codes) per word. These will
explicitly call for the execution of particular mi crocommands .
It also contains a control field which has to do with memory
accessing (allowing or inhibiting) and address incrementation.
It is sometimes maintained that the combinative properties of
this word allows a vast number of unique instructions variously
estimated at 8 to 12 thousand. Only a relatively small fraction
of these are meaningful, however, and fewer yet are useful.
Such sales arguments miss the point since the real strength
of the machine involves the way combinations of logands (the
more common ones, usually) may be put together rather than
the ability to call on an unusual or esoteric instruction from
the large number available.
The higher order interpretive language consists of a string of
logram calling sequences. The symbolic names of the lograms
are arbitrary in the sense that the programmer can name and
design his own. The assembly program will assign the starting
addresses for the corresponding logand strings.
The logram calling sequence is specified to the computer by
placing in sequencial cells the starting addresses of the
corresponding machine code subroutines. Interspersed among
these addresses are the addresses of any operands needed.
Thus, the interpretive mode code (the logram calling sequence)
consists simply of a string of addresses of subroutines and
operands. It is said that these lograms correspond to the
wired-in instructions of other computers, but a closer look
would suggest they correspond more closely simply to closed
subroutines, which, in fact, is what they are. However, a
unique method of subroutine linkage is used which obviates
2-232
the necessity for an interpretive routine to sequence the
routines. Each subroutine provides its own linkage to the
succeeding routine by accessing the address supplied in the
calling sequence and placing it in the machine program
counter. To facilitate this method, the computer (program)
maintains essentially two separate program counter registers
which indicate the current position in the calling sequence
and the program counter location within the current subroutine
The interpretive level instruction repertoire is called the
Basic Logram Package. The instructions defined by this set
resemble those of a one address computer, including single
and double precision commands. In addition, special logram
packages are available, e.g. floating-point package, matrix
arithmetic, etc. It is noted that these routines require
memory space and, in general, only those routines needed in
the application should be loaded. Although, initially,
wide varieties of instruction repertoires were anticipated,
including those which could simulate those of other machines,
in practice, the Basic Logram Set is most commonly used. In
some ways, the interpretive level is the more cumbersome.
The most attractive alternative to those familiar enough
with the machine operation is to descend to the machine code
level utilizing the more efficient methods available there.
The interpretive mode overhead tends to be constant (approxi-
mately two logands per logram) which constitutes a rather
high cost for the simpler lograms. For example, the logram
add command costs 18 microseconds (on the TRW-133), the
logand add, only 4 mi croseconds c Therefore, a combination
of the two codes is sometimes preferred, using logands for the
simpler functions (add, shift and those commands that can
utilize the efficient memory addressing available on that
2-233
level, i.e. load, store, and indexing) and using lograms for
the more complex such as sine, cosine, BCD to binary con-
version, etc. The computer, when programmed in this way,
approaches very closely the typical usage one would expect
on our conventional computer using machine code and closed
subrout i nes .
2) PB-440
The motivation for the development of the PB-440 was similar
to that held by the AN UYK-1 des i gners-the desire to develop
the capability to tailor make instructions sets specially
suited to the application. An important feature was added and
the claim was made that the first Dual Memory Stored Logic
Computer had been developed. A homogenous memory design, it
was felt, would just barely hold its own compared with con-
ventional designs (presumably because of the interpretive
mode overhead) and therefore, it would be advantageous to
place the strings of microsteps in a special module of fast
memory .
The PB-440, then, has two classes of memory; a main memory,
which operates at a five microsecond cycle time, and Logic
Memory (or "control" memory) which is a non-destructive biax
memory with a one microsecond read time. The minimum con-
figuration of the computer has 4096 and 256 24-bit words of
these types of memory, respectively.
The relatively small amount of fast memory was sufficient
to define certain basic instruction sets which could be
modified or replaced by reading in new ones, and it was
expandable in 256 word modules, if desired.
2-234
The format of the PB-440 allows two micro-orders, sometimes
called micro-steps, per instruction word. Any of 64 separate
micro-order codes may be specified. In addition, the format
contains two modifier fields per micro-order which will
ordinarily indicate one of seven working registers as operand
source and/or destination. Here, as in the TRW machines, the
working registers are available to program manipulation (i.e.
accessible to the programmer) to an unusually large degree.
The routines stored in fast memory are called microutines.
They are called into play by the "control sequence" which
utilizes special instructions designed for the purpose. It
was noted that the higher language level operation code for
the TRW machines was, in fact, simply an address; the address
in core memory of the start of the logram. In the PB-440 the
operation code will ordinarily refer to a microutine by
number (i.e. 1 of 64), and utilize a jump table to facilitate
rapid micro- i nstruct ion i nterpretat ion^' .
Special instruction sets include a systems-oriented command
list, a scientific/engineering problem-oriented command list,
and a FORTRAN set. These are interchangeable by computer
memory loading. The instruction sets are normally stored in
fast memory but are also executable from main memory. The
micro orders are tailored to recognize various data formats
such as floating point, or sign-magnitude numbers, and
alphanumeric characters.
Program optimization involves the utilization of the time
between main memory accesses, referred to as "shadow time",
during which useful computing may be accomplished (as long
as it does not involve further main memory access).
From a hardware standpoint, two-level programming is a fiction, since
the computer will always remain on one level (i.e. the lower). The
interpretive level (which Is sometimes referred to as pseudo-code)
consists of a macro- i nstruct ion control sequence which simply specifies
which subroutines are to operate. This is true of all the computers
discussed In this report.
2-235
3) C-840]
The dual memory concept is also implemented in the Collins
Computer, the C-8401 Data Processor. The microprograms are
stored in the fast memory, called the instruction memory.
This memory is composed of 1024 36 bit words with a read time
of one microsecond. The formats of the instruction word
allows either two or three transfers to occur. The transfers
relate to the exchange of information between exchange registers,
certain of which are associated with logical or arithmetic
functions. Each of the basic operations is associated with
a particular register. A large number of working registers
are thus made available to the programmer.
The main memory is composed of kOSS (expandable in 4K modules)
16 bit words with a five microsecond cycle. Thus, up to 15
mi cro- instruct ions can be performed during each main memory
cycle time. Notice that this assumes a mi cro- instruct ion to
be a part (or field) of a computer instruction word.
Macro- instruct ions are stored in main memory and constitute
a higher level problem-oriented language. The interpretive
mode linkage is effected by an interpretive routine called
RN I (Read Next Instruction). This routine maintains an
address counter which is stored In one of the exchange
registers. It also is able to provide branching in the
instruction memory to the subsequent micro-programming to
be performed.
The C-8401 was designed primarily as a communication network
processor. One of the distinctive features of the machine is
the ability to control I/O operations from many sources at the same
time. Although this is not unique in modern computers, it
IS facilitated to an unusual degree by the computer design
2-236
which incorporates the external exchange registers for this
purpose^ Input and output is specified by separate micro-
programs selected by the RN I microprogram in the same manner
that other microprograms are activated. The machine was
designed with a single application in mind but is suitable
for other applications by software modification. This
conforms to the basic rationale of the stored logic principle.
2.8.6.1.3 Characteristics
To the general observation that programming may be undertaken on a
lower level of abstraction on stored logic computers, using generally
smaller logical elements, we shall add certain other characteristics of
the stored logic implementations to date.
1) Multi-level programming - Interpretive operation is featured
on each of the computers discussed. Although complete programs
may be prepared on the machine code level, the machines are
specifically designed to facilitate subrout 1 n i zat ion .
2) Accessible Working Registers - The internal registers of the
machine are available for minute manipulation involving
transfers, temporary storage, addressing, as well as arithmetic
computation and control «
3) Adaptive instruction sets - All claim the feasibility of
custom made instruction sets to suit individual applications.
k) Relatively few clock pulses per computer instruction.
It is noted that there tends to be fewer clock pulses per machine
instruction. Usually, however, the clock pulse contains several
micro-operations and the compounding of useful functions is considered
a design advantage. This doubling up of logical operations is seen
in the fact that all three computers allow at least two command
functions per computer word.
2-237
The PB-440 and the C-8401 have in common the utilization of fast
control memory modules for the storage of the stored logic routines.
It should be noted that the development of fast (control) memory modules
for special purposes is not a unique stored logic feature. It is a
common characteristic in recent entries in the computer field.
And, finally, these machines have the common characteristic of
admitting to being stored logic computerso The implementation of micro-
programming techniques is more widespread than the number of machines
which admit to being stored logic would indicate. Stored logic is now
claimed only by those manufacturers who are committed to it. No longer
can much benefit be derived from claiming it as an innovationo If the
term stored logic does not survive, it will probably be because of
semantic difficulties and the current uncertainty among computer manu-
facturers as to whether the designation has positive or negative sales
value. The term microprogramming Is somewhat more acceptable currently
and probably more descriptive.
The long heralded 360 series announcement alluded to the fact that
the microprogramming technique is a part of the design philosophy.
In this instance the lower end of the line qualifies as being micro-
programmed. However, there is no suggestion that custom made
Instruction sets are anticipated. On the contrary, instruction
compatibility and standardization is stressed. It is Interesting
to note that several other manufacturers have indicated willingness
to conform to this new Instruction repertoire and, In at least one
case, the translation will be achieved with stored logic techniques,
2.8.6.1.4 Evaluation
The advantages and disadvantages of stored logic as a design
principle are difficult to weigh. It could be argued that the stored
logic design has not been the most compelling reason for success or
2-238
failure of those computers which have used it; nor even the most
important feature. In any case, commercial success is not a valid
indicator of design excellence since the two seem to correlate only
casual ly .
The advantages are perhaps most often summed up in the word
"flexibility"; flexibility in the sense that varieties of instructions
may be produced--that there is a selection of programming methods--that
the instruction repertoire may be changed by reading in a new set of
microstepSc Whole "logical" computers may be designed to suit particular
problem requirements; other computer repertoires may be simulated to
retain software investments; and special instructions can be designed
as needed and added to the growing library of routines.
Stored logic appears to offer certain cost savings to the
manufacturer. The less complicated control logic, the lower number
and types of components, together with the opportunities for standard-
ization of component modules make it intrinsically an attractive design.
Another advantage which was cited by early writers is that the
order code may be changed late in the development of a new machine.
And, of course, the interpretive language can be modified even after
it is built. This reflects the early concern regarding rapidly
changing instruction repertoires. Thus, stored logic was seen as a
way of delaying obsolescence. Currently, however, there is a
tendency to perpetuate code structures, at least among families of
computers; in order to maintain compatibility.
Although the interpretive operation is considered the primary
programming method, the lower level machine code is sometimes preferred.
This is occasionally necessary to exact maximum efficiency for critical
computations. Sometimes, it is found necessary to code on that level
for competitive reasons.
2-239
The interpretive overhead cost must also be weighed in terms of
storage as well as execution time. For, in the case of stored logic
computers, the (interpretive) instruction repertoire must be stored.
In the case of the dual memory machines this storage is quite expensive.
This argument may be turned around, however. The instruction repertoire
saves storage in the sense that subroutines save storage. And the
dual memory machines assuredly have a compelling reason for storing
the instruction set in fast memory.
The primary objection to stored logic computers is that it is
difficult to program them. It is felt that the logical complications
that must be dealt with are enormous; that the programmer should not
only have a thorough understanding of the subcommands, but should
have a knowledge of the Internal logic of the computer and even the
circuitry Involved. It Is said that microprogramming is not intended
for the casual user.
It is probable that the actual difficulty of programming stored
logic computers Is exaggerated. Although it takes a little longer to
develop facility at the lower language level, programmers experienced
in microprogramming are usually enthusiastic. They consider it
challenging and sometimes it takes on the characteristic of an
Intellectual recreation. However, what Is often overlooked among
those who (modestly) insist Its a "snap", is that, while it may not
seem more difficult to them, it will very likely take considerably
longer to write a string of code using microprogramming than It
would with conventional code. The apparently greater latitude to
compose elegant code even when machine efficiency results, can
sometimes turn out to be a false economy In terms of work accomplished
per unit cost.
2-240
However, It is argued that once the software is developed to
provide the desired interpretive instruction repertoire, programming
is as easy as for any other computerc The theory is that a small
team of mi croprogrammers (perhaps one) may serve to prepare all
special instructions that may be required, and will generate new
repertoires as the need is seen. This is a perfectly valid point of
view, and the capability to tailor-make an instruction set is certainly
one of the most powerful arguments for microprogrammi ng o Unfortunately,
however, this kind of software has turned out to be extremely expensive.
This is due not only to the cost of the programming effort but also
to the concomitant costs of library maintenance, documentation and
attendant activities associated with user's groups and software
development generally.
It is usually found more expedient to use combinations of in-
structions already available than to develop new ones that are more
efficient. Private Instructions make the rounds unofficially (usually
to avoid the bother of getting them accepted as a part of the library),
and standard usages become difficult to maintain. The concept of
private order codes for variable machines does present some procedural
problems. Too much flexibility may be a disadvantage.
The consensus among those who are familiar with the cost trade-offs
involved tends to suggest (reluctantly) that stored logic as practiced
by the programmer is not paying its way. A partial solution Is not to
allow the applications programmer the discretion of instruction repertoire
alteration. He is presented with a "logical" machine that is unalterable.
This unfortunately tends to negate the basic advantage of microprogramming,
i.e., flexibility. Another approach is to fix the stored logic (a con-
tradiction In terms?) in the machine. At least one manufacturer is
using this approach, and has, in effect, a plugboard type of stored
logic modularity.
l-2k\
The stored logic design should be evaluated in the context of
various other developments affecting its utility. One important
consideration has been the changing size both logical and physical,
of hardware components. Initially stored logic received considerable
Impetus from the fact that hardware replacement of large component
modules was expensive and sometimes difficult. Use of smaller
logical elements was found to be more economical. Lower component
count, standardization of pluggable replacements, and savings In
control logic; these factors were all felt to be especially compatible
with the stored logic design^
A counter trend seems now in effect which suggests that larger and
more complex logical components may be produced now at a fraction of
earWev costs. This, together with the tendency towards miniaturization,
may limit the degree of divisibility that is economical. As the cost
of hardware components decreases, the motivation for small micro-logics
may be expected to diminish. In this connection, it is noted that
the ratio between hardware and software costs is changing. There
is little evidence that the latter may be reduced In the same dramatic
fashion as the former.
With the increasing cost of software, utilization of existing
software inventories becomes very important. One approach to this
problem Is the development of translators to allow programs written
in the language of one computer to be executable on another. Stored
logic machines are very amenable to this type of implementation and, as
the "host" computer, wl 1 1 not pay so severe an execution time penalty
ordinarily as would one with a conventional design.
Perhaps in the larger view, it Is not too significant if the
stored logic concept Is maintained as an entity (although the term
microprogramming is almost certain to continue to be applied to whatever
It seems convenient and appropriate). The advances associated with
this development; language flexibility by sensitive manipulation of
small logical elements, standardization of hardware components, dual
or multiple memories to suit varying computational demands, and
translations of language repertoires to utilize software Inventories,
are likely to be of an enduring nature In computer technology;
2-242
The concept has diverged in development and has been diluted in
implementation. It has turned out to be a variation rather than a
radical departure from conventional design. Investigation in this
direction is incomplete, however, and the techniques involved are
certainly worthy of continuing study in consideration of the larger
topic of computer organization. Perhaps the development will go full
circle with a new look at Wilkes control matrlceSo
2.8«6c2 Highly Parallel Computers
A new area in computer organization which represents an almost
complete departure from conventional computer design is "highly
parallel computers'^ In these computers the arithmetic and control
logic is essentially decentralized to the extent that they exist at
nodes of a network. All of the arithmetic and control units at the
nodes then work in parallel to provide, in theory, a high speed
operation. These computers are best represented by the Solomon computer
developed at Westinghouse under contract with RADCc
There is an important question of whether these highly parallel
computers have any p^ace at all in future ANTACCS. This question
will be exami ned .
2.8.7 Analog/Digital Hybrids
It Is intended here to discuss briefly the subject of analog/digital
hybrid computers. More specifically, the current uses and possible
future uses in ANTACCS of these systems will be analyzed. Although
this work will not be a thorough examination of this type of computer
system it will nevertheless present useful information and some
opinions about future ANTACCS uses.
2.8.8 Existing NTDS Computers
To best understand what the characteristics might be of future
NTDS computers it Is necessary to understand and develop a critique of
existing NTDS computers. This will be done In this section of Technology
z-z^^
The following computers will be discussed by the completion of the
study effort: (1-20, Q-20B, 1218, AN/UYK-1, CP-667, MTDS and AIDS computers
Rather than a thorough and exhaustive presentation or summary of the
characteristics of these computers, there will be a short critique stating
the more and less desirable aspects of each of them.
2.8.8.1 Introduction
The current and widespread employment of the Q.-20 computers by the
Navy for many critical tasks makes them a vital component of the Navy's
data processing capability. With this proportion of the Navy computing
effort centered in (i-20 computers, and with the Q-20 compatibility be-
ing built into the CP 667, it Is necessary for the ANTACCS study to
investigate in some detail the capabilities and limitations of the Q-20A
and Q.-20B. The effort is approximately 50% completed at this time and
a technical report will be published as the effort continues.
The requirement of the armed forces to meet the exigencies of modern
warfare has led to the development of computers especially designed for
military environments. These computers are the chief processing elements
in systems designed to have very rapid response time and sufficient
reliability for both defense and attack situations. The best-known
computer in this class is the NTDS unit computer Q.-20. All the military
computers, in addition to having the obvious capability of being opera-
tional in severe physical environments, must have the additional qualities
of high reliability, low maintenance requirements and complete engineer-
ing documentation.
To increase their capability for field employment, the NTDS computer
(and the Army Fieldata computers) have adopted standard strobe philosophy
for interfacing peripheral equipment. The Navy version is labeled the
"NTDS Interface" and is included in all equipment that Is intended for
Navy employment. It is the signature of NTDS. The use of the NTDS
Interface permits peripheral equipment to be added or removed from a
computer system in the field by simply plugging or unplugging them.
2-244
These military characteristics, when combined with their general
computer characteristics, have made these computers candidates for use in
other military systems; and (in their commercial counterparts) candidates
for use in some non-military applications.
The list of military computers with the above characteristics has
become relatively large. The Army Fieldata program included IBM's
IMPAC, Sylvania's MOBIDiC, Phiico's BASiCPAC and COMPAC, and RCA's
MiCROPAC (the FADAC computer used by the artillery was not a member
of this family). Of these computers, only the MOBIDIC and BASICPAC
were produced in significant num.bers.
The Navy computers have been produced chiefly by UNIVAC, and include
the military computers Q.- 1 7 ("Countess"), CP 642A/US(l-20 (V) , CP 642B/USQ.-
20 (V) , and the CP 557. In addition, two smaller computers have been
built with "NTDS Interfaces": the TRVJ AN/UYK- 1 and UNIVAC's model 1218.
The only computer having the official iMTDS designation is the
CP 542A/USQ.-20 (V) which is the unit computer of NTDS. UNIVAC model
1212 or CP 542B/USQ.-20(\/) is a current, up-dated version that is,
from the standpoint of instructions, almost completely identical to
the Q.-20A. The CP 557 is a new computer that is equivalent to the
Q.-2CB, in one mode, and is a new, more powerful computer in another
mode. For the purpose of the evaluations to be published, the Q.-20
will be used as a basis for comparison.
2-266
2.8.9 Requirements and Applications of Computers in the ANTACCS Environment
In this section the needs of ANTACCS with respect to computers will
be discussed. Emphasis will be placed on the kinds of capability
necessary, the locations of the computers and the organization of the
shipboard equipments from a total systems point of vieWo Some of the
topics to be considered are as follows: computer location, inter-
computer communication, computer functions, computer system expansibility,
and shipboard vulnerability.
2.8.10 Recommendations and Comments for ANTACCS Computers
One of the principal outputs of this section will be a description
of a possible future family of NTDS computers o Items to be covered are
memory, speed, organization, circuitry and packaging, and compatibility
of these systems. This description will constitute additional input
in arriving at the decision as to whether a totally new family of NTDS
equipments should be developed or whether there should be a continual
upgrading of present NTDS computers on a program-compatible basis.
As well as the description of a possible future NTDS family of
computers and the motivations for such a family, there will be a discussion
of possible other computer-like equipments which might find application
in ANTACCS. This will include systems such as highly-parallel computers
of the Solomon type, analog/digital hybrid, and digital moduleso
2-267
REFERENCES: COMPUTER SYSTEM ORGANIZATION; SECTION 2-8
1 Wilkes, M. \l . , "The Best Way to Design an Automatic Calculating
Machine," Manchester University Computer Inaugural Conference ,
Proceed i ng , July 1951.
2 Wilkes, M. V . , and Stringer, J. B., "Micro-programming and the
Design of the Control Circuits in an Electronic Digital Computer,
" Proceedings of the Cambridge Philosophical Society ',' April 1953.
3 Wilkes, M. V . , "Mi croprogrammi ng, " Proco EJC Dec. 3-5, 1958, pp 18-20 .
k Mercer, Robert J., "Micro-Programming," Journal of the Association
for Computing Machinery , April 1957°
5 Amdahl, Lowell, "Microprogramming and Stored Logic," Datamat ion ,
Febo, 1964 pp. 24-26 o
6 McGee, W. C, "The TRW-133 Computer," Datamation , Feb., 1964, pp„ 27-29
7 Boutvvell, E. 0., "The PB-400 Computer," Datamation , Feb., 1964,
pp. 30-32o
P Beck, L. and Keeler, Fo, "The C-8401," Datamation , Feb „ 1964,
pp. 33-35»
9 Hill, Richard He, "Stored Logic Programming and Applications,"
Datamation , Feb c , 1964, pp. 36-39-
10 Buland, R. N., Baum, Ho, Real Time Impact Prediction Program,
Aerospace Science Report No. 501 UNIVAC, San Diego, 1962.
11 UNIVAC Publication MO 2762, UNIVAC 1206 Military Computer, General
Description and Input/Output Specifications.
2-268
2.9 PROGRAMMING
Much work remains to be done in the programming area, however,
an outline has been developed, and a comparison of the effectiveness
of compilers for the ANTACCS application has been undertaken. The
following is an outline of the document which is being prepared.
2.9cl Classification of Programming
The meaning of the phrases computer program, computer programmer
and computer programming is slowly changing. This section will present
a definition of the words and list the factors causing the changes o
2.9.1 .1 Definition
A historical definition of computer programs, computer programmer,
and computer programming.
2c9clc2 Factors
Factors causing a change in what constitutes a program, how a
program is generated, and who generates it.
2o9clo2cl Hardware
A discussion of the programming implications of the development
of inexpensive mass memories and display devices, the extensive use of
communications media, the effect of novel organizations of computer
hardware.
2.9.1c2.2 Systems
Developments in both ANTACCS oriented and non-ANTACCS oriented
computer systems - a discussion of the implications of the development
of communication-based data processing, information retrieval, process
control, management systems.
2-269
2.9^1.2.3 Education
A discussion of the changes in education, training and function
of programmers «
2c9o2 Sources of Information
2.9-2.1 People and Companies
A visit will be made to NEL to get a fix on their procedures.
2.9.2»2 Literature
There are many treatises on computer applications and program
results, but relatively few on programming itself » References will
be 1 i sted .
2.9»3 Programming Characteristics for ANTACCS
In this section are defined programming techniques and concepts
which have an application to ANTACCS, together with their application
The subjects covered will be those named in section 2. 9.5
2c9.4 Status
A definition of the state-of-the-art for the techniques of
program generation, program checkout, program maintenance, and the
integration of programs into systems.
2.9c4<,l Programming Tools - Off-Line Systems
2.9«^«1«1 Assemblers and Compilers
An evaluation, description, and/or comparison of assemblers and
compilers and their application to ANTACCS, namely,
1) Assemblers
2) ALGOL
3) COBOL
k) FORTRAN
5) CS-1
2-270
6) NELIAC
7) JOVIAL
8) TABSOL
2.9.^«l-2 Systems Packages - Off-Line Executives
A description of the functions, usage, and application for the
ANTACCS of:
1) Monitors
2) Data and Report Generators
3) Diagnostics
k) Debugging Packages
5) S imul at ions
2<,9.^»2 Executive Programs - On-Line Systems
A comparison of the capabilities of the executive programs of the
following systems:
1)
SAGE
STC
2)
NTDS
D-825
3)
MAC
4)
SABRE
in terms of:
2«9«^.2.1 Executive Programs
The programs required to coordinate the events which must be
processed by the systemc
2.9<.^»2.2 Time Sharing - (The Executive Only)
A comparison of the techniques of achieving a more optimum
utilization of the computing modules. To be included are discussions
of control by:
1) Highly parallel computers
2) Multiprocessing
3) Multiprogramming
2-27
2.9.^.2«3 Intersystem Control and Communication Programs
A comparison of the system components which control the operation
of and flow of data through a system. To be included are a discussion
of the requirements and techniques for:
1 ) Schedu 1 i ng
2) Buffering
3) Switching
k) Intermodule communication
2c9c^.2.4 Reliability, Malfunction, and System Readiness
A comparison of the programs and programming techniques which
can influence system reliability, system malfunction detection and
correction, and the determination of system status. Included will be
discussions of:
1) Component Diagnostics
2) Component Utilization in Programs
3) FIX
k) Graceful Degradation
5) System Readiness
2«9«^«3 System Performance
A comparison of the ways in which systems which include programmed
computers can be used to provide performance data during operation.
2c9-^.^ Program Documentation
2 .3 .k.ko] Types of Program Documentation
A description of the various ways in which programs have been
documented in the past. The description will cover for each type of
documentation, the format, the content, the relationship to other
tasks, the timing, and the distribution. The classifications are
defined to be:
2-272
1) Task description - the types of documentation which relate
programs to systems.
2) Function descriptions - the types of documentation which
relate programs to system functions^
3) Flow Charts - the types of documentation which relate programs
to hardware.
k) Comments - the ways in which documentation has been made an
integral part of program preparation.
2. 9. ^-^"2 Uses of Program Documentation
A description by type of documentation of their use in the various
phases of system implementation. To be included is a discussion of the
requirements for documentation and the purposes the types of documentation
were meant to serve, served, or could serve. The classifications to be
used are:
1) System Design
2) System installation and Checkout
3) System Operation and Maintenance
4) System Modification or Improvement
2c9»^.^'.3 Techniques of Program Documentation
A description of some machine processes, aside from compilation
and assembly, which may be used to produce program documentation
explicitly or implicitlyc To be covered are:
1 ) S imulat ion
2) Documentation Retrieval
3) Flow Charting by Computer
2.9.5 Programming Availability in the 1970-1980 Period
The section will include a prediction by subject of 2.9.^ of status
in the 1970-1980 period., In addition, it will include a discussion
of some techniques, trends, and concepts which will become important
2-273
to the programmer:
1) Logical languages - the development of pseudolanguages based
on symbolic logic and language analysis^
2) Procedural languages - the development of Algol-like languages
3) Non-Procedural languages - the TABSOL-like languages.
k) Implicit programming - techniques which replace the
programming function.
5) List Processing - techniques for information distillation
and man i pu lat ion .
6) Man-machine symbiosis - techniques of integrating procedures
and processes .
7) System Analysis - the integration of programming and the
disiplines associated with system implementation.
2.9.6 Limitations of Present and Planned Programming
A documentation of the differences between the anticipated status
of programming in the 1970-1980 period and the status of programming
nowc
2c9.7 Recommended Developments to Meet ANTACCS Needs
A documentation of procedures for eliminating the deficiencies
found in 2.9-6.
2.9o8 Evaluation Criteria Recommended
2.9.8cl Program Development
A discussion of the techniques which have been used to control
the scheduling, cost, required effort, and quality of programming for
large systems. To be included are a discussion of:
1) The Milestone System
2) PERT
3) Approaches to Cost Analysis
k) Methods of Obtaining Execution Efficiency
2-274
2.9.8.2 Program Design
A discussion of ways in which programs are designed and of the
factors which have an influence on their design. To be included is
a discussion of the factors which preclude optimum design. The
following subjects are to be discussed.
1 ) System Anal ys i s
2) Program and System Documentation
3) Computer Languages
k) Subroutines and Segmentation
5) Macro and Micro Program Development
6) Program Packages
2.9.9 Conclusions and Recommendations
This section will be completed after a thorough analysis of ANTACCS
programming operationSc
Sol
3. METHODOLOGY
3ol INTRODUCTION
This section of the mid-term report presents the material developed
thus far in the Methodology part of the studyo The material presented
is preliminary in that it will be supplemented and amplified as appropriate
during the balance of the study. As in all technical studies, final
documentation lags somewhat behind the completion of the worko One must
also consider that although this is mid-point in the study calendar, one
half of the hours budgeted for Methodology have not yet been applied.
For these two reasons, the following material is less than half of that
ultimately to be provided by Methodologyo
This section of the mid-term report is organized into three sub-
sections:
1) General Methodology
2) Implementation Methodology
3) Specific Methodology
Bibliographical material is included in a common bibliography section at
the end of the report.
In general, the work in Methodology is abreast or ahead of schedule.
3-2
3.2 GENERAL METHODOLOGY
3.2. 1 General
This area of the study is concerned with the investigation of a
few of the most important tools available for Naval Systems Planners.
These tools are called "General" since they are not directed toward
the solution of specific problems. Rather, they are tools which may
be used in any phase of system planning, and are in that sense "general"
methods. The effort, according to plan, is divided among four areas:
1) Simulation Languages
2) Techniques of Simulation
3) Mathematical Modeling
4) Critique of General Methodology
In this report, material is presented from areas 1 and 2.
3.2.2 Simulation Languages
3,2.2 . 1 Introducti on
Simulation languages are those higher order programming languages
which are especially designed to facilitate the programming, coding and
checkout of digital computer simulations.
Simulation languages (or sim languages) allow the simulator to proceed
at a greater speed in the design and construction of a simulation since they
provide for the creation of routine procedures control, and the recording
of data. Most sim languages were created for a specific purpose and have
since been expanded to treat a larger class of problems.
Simulations and models have been coded for digital computers from the
very beginning of the computer era.
The earliest of these simulations were coded using octal and even
binary absolute techniques, and fine simulations may still be produced
using machine language or combinations of machine language and Fortran, or
Algol. The use of a simulation language is in no sense required for the
3-3
production of a good simulation program. However, the use of a proper
sim language facilitates the production of the sim program and makes
the designer's task an easier one, as well as improving the speed of his
progress .
3.2.2.2 How Sim Languages Work
The construction of simulations involves the creation of lists of
things, people or events. These lists present one person (thing or event)
at a tinie to be served or operated upon by the logic of one of the central
models of the simulation.
Consider, for instance, an application of simulation of supermarket
operations. People waiting in line to pay at a supermarket would be in
a list, the head of the list to be served next by the simulation's model
of what the cash register operator does. The lists and operations may
be cascaded to show how the customer must wait at the meat counter, and in
the parking lot, or how he is served in succession by the cash register
operator and then the bag boy.
In some simulations, the lists may be few and very long. In others,
many short lists or mixtures of long and short lists are required.
In each simulation, there is at least one operation which serves the
items waiting in the lists. In complex simulations it is often necessary
to model many operations which are used to serve lists and in turn add the
just-served items to one or more other lists which will, in their turn,
be served.
The construction of these complicated models is simplified by the use
of sim languages since they provide conventions (or regularized shorthand)
for specifying the creation of lists, the operation of the serving models,
the inter-dependencies of serving models, the influence of time or other
environmental circumstances (such as tide) , etc.. Each one of the various
sim languages uses different conventions and they vary in their simplicity,
power, and general applicability. This is due primarily to the fact that
3-4
they were all created for specific purposes. Most of them have been
expanded in scope since their inception, but the prospective user of
sim languages stands to benefit if he picks a language which was originally
designed for a problem similar to the one he currently faces.
The advantage of using a sim language tailored for an application
similar to the one at hand must be weighed against the difficulties of
learning a new sim language or having a computer available for which the
language was written. But in general, one should stay as close to the
original area as possible to avoid encountering too many of the inherent
limitations which are present in all sim languages. The more complex and
complete languages may be used to simulate simple relationships and occur-
rances, but they are often much too ponderous for such use.
The use of a sim language is a multi-step operation and will be explained
here simply:
1) Develop the rules for processing the lists. These may be
mathematical models or stochastic models (models based upon
probabilities) or combinations of these.
2) Develop the rules for the creation of the lists and for items
entering and leaving the lists other than by being served by
the primary models.
3) Develop the relationships and linkages which relate the lists and
models to each other.
4) Develop the timing and operational considerations for the execution
of the simulation.
The user now begins to write the simulation using the conventions of
the language chosen. When this step is completed, the computer and the
T
simulation assembly program process the sim conventions. he output of
this step is a program (or deck) in computer language. This deck may be In
machine code such as FAP, but more often the result of the first pass is
a compiler language such as Fortran, This, in turn, must be compiled into
3-5
machine code and then converted into the binary deck which is finally
operated. This operation is the simulation being cycled. The answers
and statistical data are recorded and printed out during and following
th i s th i rd pass.
Some sim languages permit the use of machine code and compiler or
assembly language in the original writing of the simulation. This process
is called "enrichment" and enhances the basic capability of the sim
language. Enrichment permits the simulation designer to code certain
intricate portions of his simulation in machine or assembly language and
bypass various shortcomings of the sim language. Since all possible
simulation requirements cannot be provided for in a sim language, enrich-
ment capability is a highly desirable characteristic, although one good
language (GPSS) does not posess it.
The easier the sim language is to learn and use, the more stylized
it tends to be and the more limited it tends to be in terms of the flexi-
bility of what it can describe. The more capable a sim language is, the
more complex its rules are and the more difficult it is to use properly.
3.2.2.3 Some Simulation Languages
This section discusses several of the better known sim languages
as well as three which are of considerable interest to simulators at the
present time. This material will be added to substantially during the
next portion of the study. Only a brief glimpse is provided of the
languages, but it should serve to show the primary capability and application
of each .
1) GPSS II
This program is an IBM product and an outgrowth of the
"Gordon Simulator" which first appeared as an AIEE conference
paper in August 1960. GPSS I followed and the current GPSS II
is an enhanced and more flexible version of its model I
3-6
Predecessor. GPSS is designed to handle simulations of
communications systems and computer systems, where there may
be many lists of varying lengths, but where the central model (s)
of the simulation are relatively simple logically. All of the
relationships and types of operations are rigidly specified,
and GPSS cannot be enriched with any assembly or machine code.
GPSS II is available for the IBM 7090-7094.
2) GASP
GASP was developed by the U.S. Steel Corporation to simulate
the operations in various shops of steel mills. In these
simulations, the lists may be somewhat shorter than with communi-
cations simulations. At the same time, the models of the
simulation can be very complicated. GASP is one of the earliest
of the powerful, flexible sim languages. GASP permits the use
of Fortran for enrichment and is compatible with Fortran
diagnostic tools. It is available for the IBM 1620, 7070-7074,
7090-7094, and CDC G-20.
3) CLP
The Cornell List Processor, or CLP, was developed by the Indus-
trial Engineering Department of Cornell University. Its purpose
was to provide engineering students with a general purpose
simulation language that could be learned and used well in
one semester. CLP is simple in its syntactic construction
and, therefore, easy to learn. It is not completely stylized
and so, has fair flexibility. CLP may be enriched by employing
CORC statements. (CORC is a compiler language written by Cornel
with an eye toward ease of use). CLP combines simplicity with
flexibility and enrichment. It is availably only for the
CDC 1604.
3-7
4) DYNAMO
DYNAMO is a capable sim language but one designed expressly for the
construction of simulations employing differential equation models.
It was developed at the Massachusetts Institute of Technology for
the IBM 7090-7094. It is an interesting and valuable engineering
tool but one which has rather limited scope of application.
5) CSL
Control and Simulation Language (CSL) was developed by IBM (UK)
in conjunction with Esso (UK) for the purpose of simulating
corporate operational problems of large scope. These are problems
such as the operation of a port-tank f orm-- ref i nery complex
receiving crude oil by tanker and shipping output by rail, truck
and barge. The real capability of CSL lies not in the creation
of long lists, but in the ability to create and manipulate many
complex operational models and to cascade these models in
tremendously complex ways.
As might be expected, CSL has a difficult syntax and many
formidable construction rules. It may be enriched with Fortran.
It is a "three-pass" language which would ordinarily be of little
concern. For CSL, however, the first pass is made on the IBM
1401 (U.K. Model) and the last two passes on the IBM 7090-7094.
It is not now available in the U.S. nor outside of IBM (U.K.),
but it is supposed to be available generally in Britain in
late 1964.
6) SIMPAC
SIMPAC was developed by the System Development Corporation as a
research tool and is one of the more powerful of the simulation
languages. Almost by definition, this makes it one of the most
difficult to learn. SIMPAC is run on a 7090-7094, and while
other sim languages for the 7090-7094 operate under Fortran
control, SIMPAC normally uses SOS control and this requires 14
3-8
tape drives. SIMPAC could be run on a large Fortran 7090 but
it would be cumbersome to do so. SIMPAC can be enriched, but
not with compiler language. Its only enrichment is with a
machine mnemonic code (SCAT) .
Many of these limitations are of little importance to the skilled
programmer with a very large 7090-7094 installation at his disposal.
Still, they represent substantial barriers to many potential
users.
7) SIMSCRIPT
SIMSCRIPT was developed at Rand to provide more efficient pre-
paration of simulations being used in various Rand projects.
It operates on an IBM 7090-7094 under Fortran control. It may
be enriched by Fortran statements and by code written in FAP
(a Fortran compatible mnemonic machine code). This FAP enrichment
feature gives it great capability, at the same time allowing for
enrichment by the eas ier-to-use Fortran.
SIMSCRIPT is complex In its syntax and rules, and is difficult
to learn and use well. Some of this is compensated for by its
excellent documentation which includes a pamphlet on how to cheat
the sim language grammar rules to provide even more capability.
SIMSCRIPT is the most popularly used of the powerful simulation
languages and will probably remain so for some time to come.
8) MILITRAN
MILITRAN is a military simulation language developed under
contract to the Office of Naval Research by Systems Research
Group, Incorporated. It is designed to run on the IBM 7090-7094,
and is primarily designed for the simulation of military operations.
3-9
rather than the simulation of system operations. The first
available comprehensive technical output of the MILITRAN
Project arrived during the preparation of this report and could
not be reviewed in time for inclusion. It will be covered in a
later report.
3.2.2,4 The Appl ication of Simulation Languages
The simulation programs which sim languages prepare are not quite as
efficiently coded as those a very skilled programmer could write, but they
are available in a small portion of the time that programmer would require.
In simulation programming, as in other programming, understanding the problem
and deciding what to do take impressive amounts of time. But, once that
is done, simulations may be prepared much more easily , accurately and
speedily using a sim language.
The system engineer has two major uses for a sim language. In the
process of design, he often wants to check the performance of portions of
the system or simple sets of interactions. For this purpose he wants a
quickly used, simple sim language. For this purpose CPL would be ideal,
but it is not generally used, although it could probably be made available.
He must use GPSS II or something more complex than he needs like GASP or
SIMSCRIPT. With these limitations, most system engineers do not simulate
quick and dirty problems as they would be more prone to do if CLP or a
similar simple language were available.
The second simulation requirement of the system engineer is the one of
simulating large portions of the system and finally the entire system. This
type of simulation is normally not prepared on a short term basis and the
more powerful languages SIMPAC or SIMSCRIPT can appropriately be used. CSL,
when it becomes available, will be highly desirable for these large scale
s imu lat ions.
3-10
3.2.2.5 Current Developments
More than one computer manufacturer is known to be or reported to
be preparing simulation languages. These seem to be at the most powerful
end of the capability spectrum, SOL has been developed by a group of
system engineers at Burroughs, Pasadena. I t nun s on the B-5000 and is '
extremely powerful, reportedly as capable as SIMSCRIPT or SIMPAC. In
addition, SOL may be enriched with ALGOL statements, and runs under
B-5000 ALGOL control. It is also constructed in a completely different
manner from the balance of the sim languages. It {s"syntax oriented"
which means that the compiler and its conventions more closely parallel
our natural language In operation and, therefore, the grammar and con-
struction are much easier to learn to use.
SOL was not mentioned In the previous section since It must be released
to the public through Burroughs, Detroit, and at last report they seem to
have little interest in doing so.
There are no reports of smaller scale languages .in development.
3.2.2.5 Comment
This leaves the language spectrum available to the system engineer
looking like this: 3 , ^p^^
GPSS II CLP GASP SmSCRIPT
SOL
with DYNAMO and MILITRAN each off in its own specialized but useful
dimension. GPSS II is capable but completely unchangable since It cannot
be enriched, CLP could probably be made available by private treaty, but
It Is not well known and only runs on the 1604, GASP Is old, but capable
and runs on several machines including the G-20, SIMPAC has great power
but severe limitations. CSL Is not available yet, and SOL may never be.
3-11
The choice is really GPSS II, GASP or SIMSCRIPT and with these three
languages, the simulation requirements of all phases of system engineering
may be met satisfactorily. But special applications make CLP, SOL and
CSL continue to look very promising to those who follow simulation closely,
3-12
3.2.3 Technique s of Simulation
Several papers concerning various aspects of simulation have
been planned. They are:
1) Simulation and Modeling Techniques for System Design.
2) Simulation for Command Control System Checkout .
3) Simulation for Training Purposes .
4) Simulation for Command Control System Development o
5) Simulation: A General Discussion and Survey .
6) The Role of Simulation in T est and Evaluation of Navy
Command Control Systems at Point Mugu .
7) Simulation in Real-Time for the Line Commander .
8) Design of Simulati ons of Real-Time Systems.
9) Mathematical and Physical Modeling Techniques .
Of this series, 1, 2, 4, and 6 are included in this reporto The
balance of the material will be presented in subsequent reports.
3.2c3.1 Simulation and Modeling Techniques for System Design.
3o2.3ol.l Introductory
Why does a system designer use simulation techniques? "Simulation"
and "modeling" imply imitation while "design" implies creation. Actually,
the system designer does not simulate and model for the purpose of
creating system designs but for the purpose of testing system designs.
The system designer can test and examine the early forms of his design
with simple diagrams and hand calculations. His intuition and experience
tell him that one equipment configuration is more functional than another.
However, as the design becomes more advanced, he finds it increasingly
difficult to evaluate the design trade-offso Finally, the design is too
3-13
complex. He can no longer visualize the dynamics and interrelationships
of the myriad of components in his creation.
How can he be sure his design will perform as he hopes or expects
when it is subjected to the stresses which the real world will impose?
One method would be to build a prototype system and subject i t to a
simulated real -world environment. There are obvious reasons why this is
an unrealistic approach, especially for military command and control
systems:
1) Simulated environments, such as military maneuvers, are
expensive in terms of time, manpower, and money.
2) It is difficult to reproduce real-world environments for
repetitive tests of system prototypes.
3) System prototypes are expensive and may require years of
development.
Computer simulation is a relatively fast and inexpensive alternative
method of testing system designs. Simulation, however, is limited by the
ability of the simulation designer in creating an accurate model of the
system components and their interaction. The system components may be
computer programs, human controllers, information channels, sensors,
weapon systems, etc. Each component and its dynamic relationship with
the other system components must be represented accurately to achieve a
valid system simulation.
Here lies the "fly in the ointmento" The fly is the human
componento The actions of human components are relatively unpredictable,
especially when they involve evaluation and decision processes.
There are two general classifications of system simulations:
Man-machine simulation a nd all-computer simulatlono The following sections
will discuss the application of these two types of simulation to the design
of command and control systems.
3-14
3.2.3.1 o2 Man-Machine Simulation
This section discusses operations simulation o Operations
simulation is used to simulate the operation of a command and control
system at the interface between management personnel and display devices.
Figure 3-1 illustrates a command and conLrol environment and the man-
machine interface which is simulated with an operations simulationo
An operations simulation presents simulated information to
management personnel and modifies the information appropriately depending
on their response to the information. In other words, an operations
simulation consists of:
1) Management personnel
2) Communication equipment, and
3) Information exchange
The method of controlling the information exchange between the management
personnel and the communication equipment depends on the information rate
and quantityo If only a small amount of information is communicated, the
information exchange might be Implemented manually with switches and/or
grease pencil displays. However, since the information rates and quantities
are high for command and control systems, operations simulations generally
employ computers to control the information exchange. The computer is also
used to simulate other components of the command and control environment,
i.e.; sensing devices, controlling devices and the external world
act iv i ties.
Figure 3-2 illustrates an air-traffic-control (ATC) simulation
which is an excellent example of an operations simulation. The computer
is used to simulate a portion of the external world activity (the movement
of the aircraft) and part of the command and control system (the radar
tracking inputs).
3-15
V:
External
World
Activities
Man - Machine
Interface
t--t
(^
^
"T^
Sensing
Devices
■
Controlling
Devices
1
i
/
i
Data Processing
Devices
Communication
Devices
Management Personnel
Figure 3-].
COMMAND AND CONTROL ENVIRONMENT
3-16
Pilot
ATC Display
Charactron
Display
TV
Display
Keyboard
Figure 3-2,
FUNCTIONAL DIAGRAM OF AIR-TRAFFIC-CONTROL SIMULATION
Ultimately, the objectives of system designers are to increase the
effectiveness and functionality of the system design and at the same
time, reduce the time and cost of implementation. Operations simulation
is a tool which can be used by system designers in achieving these
objectives. However, these objectives are too general to be used in
planning specific simulation runs. Each simulation run or series of runs
is designed to produce data which will be used to form specific conclusions
about the system design.
The system design is mutually determined between the system users
and the system designers. The design is arrived at by using past
3-17
experience, imagination, projection. Intuition, etc. However, many
system parameters are difficult to evaluate: the type of information
displayed, the frequency of information updating, the number of operators
required, the performance of the operators under peak loading, the
reaction time of the operators, the types of operator errors, the
consequences of operator errors, unnecessary control options, necessary
automatic modes of operation, etc. These parameters affect the system
design considerably. They affect the quantity of communication links, the
size of computer memory, the speed of the selected computer, the computer
software, etc.
The model of the system may contain hundreds of parameters which
must be examined. Sometimes it may be necessary to examine only one
parameter with a series of simulation runso For instance, the effect of
aggregate or lumped radar returns during peak loading on the commander's
actions could be tested with a series of simulation runs.
Operations simulations have been used to determine these types of
parameters at the System Development Corporationo The following paragraph
is an excerpt which describes the results of an operations simulation of
a Manual Air Defense Master Direction Centero
"By means of the bioscopic simulation technique described, It was
possible (1) to obtain measures of the rate of information processing
throughout the system, the number and kind of errors which were made, and
a measure of the performance variability of all the operators; (2) to
obtain a clear picture of the job requirements at each of the positions
throughout the entire complex in terms of the information requirements
of the posi tion; (3) to observe performance degradation under various
conditions of stress; (4) to determine how centralized supervision from
the MDC could be effectively maintained and how restricted communication
lines should be used to maximize information flow; and (5) to provide an
overall estimate of system efficiency in terms of the primary operational
3-18
mission of the MDC complex - namely, how much threat warning time could
be expected under various kinds of stress conditions, both in peace and
2
war situations."
Sometimes a system design contains latent parameters as
requirements which are illuminated during an operations simulation. In
a sense, the simulation is used as a stimulus for ideas whicli will improve
the system design. This point is expressed in Item (2) of the preceding
ci tation.
Hopefully, operations simulation will provide feedback to the
system designers for the improvement of the system design. This will
reduce the time and cost of implementation by reducing the modifications
to the production model of the system. This point is well illustrated by
the cost of computer software changes compared to the original software
in the SAGE System. In the opening address of the 1963 Fall Joint
Computer Conference, it was stated by Major General Terhune that the
cost of each computer instruction for the original SAGE software was
approximately $32, and the cost of each computer instruction for
modifications to the original software was estimated between $100 and
$l,000c
It is concluded that an operations simulation should not be
implemented unless a definite need is identifiedo In addition, a
simulation technique should be developed which will satisfy that need.
After the simulation is implemented, each simulation run should be
planned to yield results which can be used to form specific conclusions.
"The most common pitfall in simulation is the failure to
anticipate how simulation results will be used. Simulations can
produce literally mountains of data. Selection from these data,
reduction into summaries, and analyses of significance must be anticipated
and, in fact, preplanned. Perfectly good simulations have been known to
3
fail for lack of this planning."
3-19
This section describes the simulation laboratory and equipment
which is necessary to conduct operations simulationo The simulation
laboratory is the enclosure which houses the equipment and personnel.
The equipment consists of computer hardware^ computer software and
communication deviceso
1) Simulation Laboratory
The laboratory consists of many rooms, depending on the
amount of hardware required to support the simulation and
the type of environment which must be simulated. If a
decentralized command and control system must be simulated,
a room or compartment may be required for each group of
management personnel . The Command Research Laboratory at
SDC has a modular construction so it can be easily partitioned
into various sized compartments. The ceiling of the laboratory
is constructed of interchangeable squares which contain
lighting, power and communication outlets.
SDC is also conducting several man-machine simulation studies
in the System Simulation Research Laboratoryo Each study is
supported by the adjacent computing facility which is used
primarily for general data processing functions, with
occasionally scheduled simulation runso
In addition to the data collected by the computing facility
during simulations, a large amount of Information Is gathered
by observationo The observers are separated from the
participants by one-way glass. Each study has an observation
area where the simulation controllers can study the simulation
participants. The Systems Logistics Laboratory at the Rand
Corpe consists of a similar arrangement.
Simulation laboratories should be built adjacent to existing
computing facilities to take advantage of their data-processing
3-20
supporto This can reduce the equipment costS; which are
high in man-machine simulation^ by making efficient use of
computer time=
The laboratories should have a modular construction which
can be easily modified to suit varied group configurations.
This type of construction is expensive^ however^ and may not
be warranted if an extended simulation activity is not
foreseen.
One of the fringe benefits of an operations simulation is Its
system checkout capabilityo if the laboratory is large enough,
actual system hardware can be incorporated into the simulation
as it is developed. The actual system computer can be used
in the simulation when ready and the general computing facility
could be relegated to furnishing system inputso
"In support of its design and development responsibilities in
the Strategic Air Command Control System (SACCS) project, SDC
established a Simulation Facility (SIMFAC) in Paramus, New
Jersey. The SIMFAC is a physical model of the SAC Underground
Command Post complete with Command/Control personnel stations,
capabilities to produce simulated SACCS hardware printouts and
wall displays. There is a soundproof observation deck in which
SIMFAC personnel perform actions necessary to simulate all
external occurrences starting from an Intelligence buildup to
changes in threat responses. It is in this manner that many
of the operational design concepts for Command/Control function
3
have been derived and validated."
2) Computer Hardware
Large general purpose digutal computers are generally used to
control operations simulations because:
3-21
a) They have software which facilitates the development and
modification of complex simulation programs.
b) They have speed and capacity for processing the simulation
tasks in real-time, and
c) Many organizations already use a large general purpose
digital computer for their data processing worko
The Systems Simulation Research Laboratory at SDC uses a
Philco 2010 to control several man-machine simulations. The
computer is normally used for general data processing tasks
with an occasionally scheduled simulation support function.
The computer can also operate in a pseudo mul ti -programming
mode in which a data processing program can be interrupted and
saved for later completion while a simulation program is
executed.
An air-traffic-control simulation used an IBM 7090 which
was equipped with interrupt features, a real-time clock and
special interface equipment o
"Based on early experience it was decided that a large high-
speed digital computer would be required to properly conduct
these simulation experiments^ At that time the IBM 709 was
installed at the National Aviation Facilities Experimental
Center and was, therefore, recormended for use with this
Computer Driven Simulation Environment (CDSE). Subsequently,
the IBM 709 was replaced by an IBM 7090= The 7090 computer
was equipped with two data channels, a Direct Data Connection,
and a real-time digital clock. The simultaneous read-write-
compute feature of this computer made possible the transfer
of large blocks of data without excessively increasing program
3-22
execution times. The Direct Data Connection provided high-
speed data transfer to the Display Buffer system and from
the Keyboard Data Entry Systemo"
The computer was used to simulate terminal air traffic in
this simulation which illustrates one of the advantages of
using a digital computer, i.e., the capability of easily
simulating other system or real-world components.
"A digital computer can be used to generate these targets
within this framework of requirements. It permits the
inclusion of controlled navigation and speed error distribu-
tions so the effect of errors on the simulated system can be
studiedc The versatility of digital computer generated targets
also allows additional functions to be programmed in the
aircraft simulation that were unavailable in previous analog
simulation equipment, eog., the ability to realistically
navigate the aircraft using the instrument landing system
(ILS), or realistic simulation of take-off acceleration,
etco"
"Simulation of Sub-Systems o A digital computer can be used
to simulate a large variety of complex sub-systems that might
occur \n present or future air traffic control systems.
Simulation of these sub-systems avoids the necessity of
developing these often complex equipment until their value
has been reasonably well determined. Some examples of
these sub-systems might be the radar or other form of position
acquisition system, the radar or beacon tracking system, or
a beacon attitude receiving and display system."
In the last five or six years, computer speeds have increased
to the point where they can be used much more extensively for
real-time operations simulation. The most effective type of
3-23
computer is a large scale general purpose digital machine
with interrupt features, real-time clock and standard
display interface equipment.
Mul ti -programming techniques can be used to reduce the
cost of operations simulation by using computer time more
efficiently. Cost can also be reduced by making use of an
already existing computing facilityo
3) Computer Software
To date, there are no software packages specifically
tailored for the implementation of real-time man-machine
simulation programs. This is probably due to the limited
use or novelty of the technology. There are compiler
languages which can be used for real-time programming, but
generally real-time programs are seldom coded in a compiler
language because the generated code is not efficient and
real-time compiler languages are not available for most
computers. However, if computer time is not an Issue, some
efficiency could be sacrificed for the advantages of a
compiler languageo
The computer software required to support a large simulation
effort will be an expensive Investmento It will be modified
more than any of the other equipment which make up the total
facilitleso A real-time simulation programming system should
be established before any simulations are attempted. The
programming system will aid In the planning and coordination
of the simulation development. It will also facilitate program
modifications and documentation by establishing standard
procedures. In addition, new personnel will become familiar
with a wel 1 -organized and documented system with the least effort.
3-24
Once a flexible programming framework is established, the
development of the many programming segments which compose
the simulation can begino Some of the program segments
will contain actual system software logic, the other program
segments will be simulations of real-world elements^
The actual system software logic will undergo an evolution
as the system requirements solidify through use of the
operations simulation. The outgrowth of the evolution will
be a set of program specifications which fit the needs of the
final system with a minimum of modification. In a sense, the
computer program specifications are written before the
hardware Is selectedo These specifications will be helpful
In selecting proper computer and auxiliary memory units.
The simulated environment software will provide substitutes
for the real-world elements which are absent in the simulation
The software will be the implementation of mathematical
(Including logical) models which represent radar Inputs,
weapon effectiveness, threat dynamics, system errors, etc.
The speed of the computer must be adequate enough to process
the actual system software logic and the real-world models in
real-time. Consequently, time Is the limiting factor when
designing the mathematical models. For example, it may be
necessary to compute radar detection on a probability basis
using a stochastic process rather than modeling the radar
search pattern and testing to determine when the radar beam
Intercepts an alrcrafto
4) Communication Devices
Communication devices consist of displays and consoles
required for communication between the computer and the
3-25
management personnel. The equipment depends on the type of
information which must be communicated. Cathode ray tubes,
TV tubes, slide projectors, keyboards, buttons and switches
are commonly used devices. The Systems Simulation Research
Laboratory at SDC uses all these equipments in their man-
machine simulation studies. Manual display devices, such as
weather status and equipment status boards, are economical
displays which are frequently used in the early stages of
operations simulation.
"The displays chosen were of the Charactron type (19" diameter )
and had ability to display alphabetic and numeric information
(fixed character size), special symbols (for aircrafts, fixes,
etc.) as well as lineso Since the display format was under
program control, flexibility existed for simulation on these
displays for any desired situation. Initially, there were
four displays connected to the computer through the Display
Buffer systemo Three of these are presently used for air
traffic controller consoles and one provides data for the
simulator pilots by means of a closed circuit TV system.
Four high resolution cameras transmit the data from the
Charactron to six simulator pilots' TV monitor displays^"
"A set of input keyboards are required to allow controllers
and pilots to communicate with the computero Six keyboards
are used for simulator-pilot input functions and three for
the air traffic controller positions."
General purpose display equipment should be used for operations
simulation during the design phase of system development. This
will enable the equipment to be reconfigured in order to test
a variety of operating modes and display configurations.
3-26
"Previous experience has shown that buffered general purpose
displays are needed for controller and pilot situation
displayso These units should also be capable of displaying
tables and other alphanumeric information to the controller.
By building the simulation around a general purpose digital
computer the flexibility of a programmed display system with
no "built in" format restriction can be readily obtained."
After firm communication requirements have been determined,
more elaborate consoles and displays may be constructed in
order to refine the system design.
5) Additional Hardware Equipment
Additional equipment may be warranted to achieve more
realism. It might be necessary to make models or photographs
of terrain which are scanned by closed circuit TV cameras to
realistically display military developments.
The computer which Is used for the simulation must process
the actual system software and also simulation models. If
the system software becomes elaborate, the computer may not
be able to compute both in real-timeo In this case, it may
be necessary to use another computer to process the simulation
models. This computer could supply all the inputs to and
receive all the outputs from the computer which executes the
system software logic.
Hybrid simulation techniques can also be used to relieve the
digital computer of burdensome equation solving. An analog
computer might be employed to simulate an entire air battle
involving many interceptors and threatSo
Imagination is the only limit to the degree in which realism
may be achieved in operations simulation. Some man-machine
simulations no doubt employ sound effects to increase the
realism of the simulation.
3-27
Operations simulations are conducted using a "gaming"
approach. A threat model is designed which will present
a situation to the management personnel through their
communication devices. The simulation will respond to
the actions of the management personnel by displaying to
them the consequences of their actions. The goal of the
management personnel is to "defeat" the threat model.
An extension of this technique induces a note of competition
into the simulation by using two teams:
A "red" team which Is completely familiar with the system and
simulation, and "blue" team which is composed of system
designers and system operators.
The red team designs threat models and tactics which it feels
can best challenge the system. The simulation should be
designed to provide the red team with as much flexibility
as possible. This could be achieved by allowing them to
write software models which would be included in the simulation.
These models could include elaborate tactics which could be
changed dynamically depending on the response of the system.
Of course, this technique must use referees to monitor the
red team's threat design to insure it does not violate any
rules of the game. The referees also test the simulation to
see if it operates correctly prior to the simulation run.
After the simulation is checked out, the blue team is briefed
and they "man the consoles" and do their best against the
threat model. The object of this approach is to test the
total capability of the system and locate any weak pointSo
The following two paragraphs are from a paper which describes
2
the use of an Air Defense Master Direction Center Simulation.
3-28
"Four test crews were used at each of the experimental
sites in the MDC complex. All four crews were exercised
once each day for six days in each of the eight test
conditions for a total of 192 exerciseso The test
situations included three wartime and five peacetime STP
problemso All were high load problems designed
specifically to stress weak points in the complex."
"Data were collected regarding the time required for the
air picture to be displayed on the MDC vertical board.
Measures of timeliness^ accuracy, and completeness of
information were used as criteria for the evaluation of
the efficiency of the system under the eight test
conf igurat ionso"
3.2o3olo3 All-Computer Simulation
This part of the paper will use a hypothetical system design to
illustrate how all-computer simulation can be used as a system design
tool o The discussion presents a brief description of the system design,
a system model and a simulation technique.
The mission of the system is to defend a circular area against
attack from approximately 25 ballistic missiles with an 80% probability
of destroying 25 missiles, and at 95% probability of destroying 20
missi les.
The system design consists of five interceptor complexes equally
spaced on the perimeter of the defended area. Each complex has a
computer, command personnel, interceptors, detection equipment and
tracking equipment. All of the five perimeter complexes are connected
through a master control center which monitors the entire systemo Figure
3-3 is a simple illustration of the system deployment and communication
1 inks.
3-29
Figure 3-3,
HYPOTHETICAL SYSTEM DESIGN
The following list describes some parameters of the system design
which affect the capability of the system. The task of the system
designer is to determine a set of parameters which will permit the system
to fulfill its mission. At the same time he must consider the system cost
involved with each ee lection of parameters.
1) The detection ranges of the missiles are uniformly distributed
between P, and P yards*
2) Each complex has P« interceptors which can be launched at
a maximum rate of one interceptor every P. seconds.
3-30
3) The probability; P ^, of one interceptor destroying one missile
is a known function of the position and velocity of the missile
at the launch time of the interceptor.
4) The time required for the master control center to assign a
BM to another complex is normally distributed about P^ seconds
with a known variance; the assignments are processed in order,
one at a time.
5) A peripheral complex is destroyed if a missile impacts within
Pq miles of the complex.
In addition to these types of parameters, decision rules must be
established which govern the use of the system, i.e.:
1) How many interceptors are launched at each missile?
2) How are weapon assignment conflicts resolved?
3) How much control is exercised by the master control center?
Can the system designers determine a set of parameters and
operating procedures which they feel will maximize performance and
economy, using their intuition and experience? If they determine a set
of parameters, how can they demonstrate the performance of their design
for final approval? For example, how can they show the economics of
hardened complexes vs. more destructive interceptors?
Although simulation is not by any means a panacea, it is being
used successfully to shed light on these types of questions. Some of
the evidence of this are the number of simulation languages now being
used to study the "machine-shop" class of problemso The block diagrams
or flow diagrams which describe this class of problems are very similar
to the block diagram which would be used to model the hypothetical missile-
interceptor system. The jobs in the machine shop simulation would be
analogous to the missiles, and the machines would be analogous to the
peripheral complexes.
3-31
Figure 3-4 is a simplified block diagram of the missile-interceptor
modelo Even with this simple problem^ however^ it would be difficult to
determine the capability of the model using pure intuition.
A computer program must be written to exercize the block diagram
model of the system designo If a simulation language is not used, a
tailored computer program must be written.
Many simulation runs are required to obtain results from a
stochastic system model. The approach to the missile-interceptor model
would be a Monte-Carlo technique:
1) Design a variety of attack configurations which span the
spectrum of expected threats.
2) Select a desirable set of model parameters and decision rules.
3) Run the simulation many times, possibly a thousand times,
for each attack configuration.
The effectiveness of the model against an attack configuration will
be proved if 25 missiles are destroyed in 80% of the runs and 20 missiles
are destroyed in 95% of the runs. If the effectiveness of the model is
inadequate, another set of parameters must be tested in an effort to
improve the performance. In any case, the model parameters should be
varied to study the sensitivity of the model's capability to critical
parameters, the number of interceptors launched at each missile for exampleo
The following paragraph describes a general system simulation program,
essentially a simulation language, which is used at the System Development
Corporation:
"A major contribution to the tools available for data processing
system analysis has been developed in the form of a Data Processing
System Simulator (DPSS) o The DPSS is an extremely flexible general
purpose computer program that provides system performance data on a
3-32
No
( Start V-
Yes
(Undetected ^
Status J
Yes
Yes
Set Status of
all BMs to
Undetected
Set Status ^Y
Assigned T
Put BM in
Reassignment
Queue
Initialize
List of BMs
No
Generate Position
of Next BM
Status ?
C Assigned ^ (^ Destroyed^
Status J \^ Status j
Yes
Calculate
Destruct
Probability
No
No
Yes
CSet Status^\_
Destroyed J
No
Yes
Eliminate
Reassignment
Capability
Yes
Yes.
Remove
Complex
from List
Increment
Time by
Delta Time
CSet Status^
Destroyed^
Figure 3-4o
MISSILE- INTERCEPTOR MODEL
3-33
proposed new design or modification to an existing design prior to
making equipment selections and commitments or performing any significant
computer program design. The total system design including the software
and equipment portions can be subjected to a rigorous analysis and
evaluation early in the design process so that key decisions can be made
in the areas of:
1) The kind of equipment to be used.
2) The number of each type of equipment.
3) The kind of data processing discipline and strategy required.
4) The projected performance of the system under varying loads =
5) The system's maximum capacity.
6) The system's ability to respond as a function of loading
capacity, and environment."
The degree of validity of the simulation results is dependent on
the degree of accuracy of the system model. If one component of the
system is modeled inaccurately, the system is sensitive to that component,
the results of the simulation will be misleadingo
If the system contains an active human component, it will be
difficult to develop an accurate model of the systemo However, if
acceptable models can be developed for human elements, all -computer
simulations can be used to save time and money. Models of human elements
can be developed using man-machine simulationso This can be done by
recording the reaction of many elements in response to a specific display
configuration. The recorded data can be used to establish "typical"
operator decisions, errors, etco Unfortunately, models developed in
this manner can only be used for the simulation of one situation (one
attack conf igurat ion) o
3-34
The performance of the missile-interceptor model is supported by
human components which never make mistakes when assigning missiles to
complexes. A more realistic model might occasionally cause missiles to
be assigned to complexes which have all destruction probability. In
either case^ it would be difficult to determine the performance of the
operators without the aid of a man-machine simulation.
However, even with erroneous components in the system, simulation
results can be valuable in determining the relative importance of
different components in the system. Suppose that the operators in the
BM-interceptor system made errors frequently which affected the total
performance of the system by 25 percent. Even with this error in the
model, the simulation might still be used to evaluate the trade-offs
between "hardened" complexes and more accurate interceptors.
3o2.3.1.4 Conclusions
Operations simulation and all-computer simulation both can be used
to answer system design questions. Both types of simulation have
advantages and disadvantages which will be presented in the following
di scuss i ono
Operations simulation is a valuable tool for determining the
operational requirements of a command and control system. This is
accomplished by entering actual operating personnel into the simulation
so they can uncover functional difficulties of the system design before
the system is produced and used in the field. This point is emphasized
4
in the following paragraph.
"The problem described in this paper is probably characteristic
of many large system-design problems. Certainly the need for such
operational control systems is expanding in the military, and in both
civilian and military the need for improved management systems has been
ever presento Simulation techniques can prove of much help in these
3-35
problems by providing a means of pooling and integrating knowledge from
many sources and by providing the opportunity to integrate and vary the
many variables and parameters that compose such systems. Although most
published simulation experiences have involved all-machine models^ we
have found much value in man-machine simulation when the problems have
involved organizational interactions^ the design of information systems,
and conflicting or interacting decision rules, since these undergo
considerable development during the simulation process."
The main disadvantages of operations simulation are:
1) They require elaborate hardware equipment, i.e., displays,
special interface equipment, larger facilities, etc.
2) They require more time to implement,
3) They require more time to run, i.e., they normally
run in real-time and require the briefing and debriefing
of operating personnel and
4) They require the use of a trained experienced operational
crewo
The advantages and disadvantages of operations simulation are
reversed for all-computer simulation. All-computer simulation requires
no elaborate equipment or facilities (other than a three million dollar
computer), it is relatively fast to implement (especially when written
in a simulation language) and it can run faster than real-time (if it
is not too large). However, it generally cannot be used to uncover any
functional difficulties of the operating personneK
Consequently, all-computer simulation is normally used in
conjunction with Monte-Carlo techniques. These require a great number
of simulation runs or for the evaluation of large numbers of design
alternatives. An all-computer simulation written in SIMSCRIPTwas used
at the RAND Corporation in a logistics study to evaluate a large number
3-36
of scheduling procedures. The two most optimum procedures were then
evaluated in an operations simulation of the same system.
In this way, operations simulation and all-computer simulation
can be used in conjunction with each other to solve system design
questions rapidly and economically.
3.2.3ol.5 A Recommendation
It is recommended that operations simulation and all-computer
simulation be employed at the earliest possible stage of ANTACCS design.
Actual Navy operating personnel should be used in an operations simulation
so they may evaluate the functionality of the operating procedures and
total system concepto
"Operations simulation can deal with hardware, command decisions,
Human interaction, operating procedures, situational change - in fact, all
the important factors operating In and about a system - In such a way that
inputs are identified, performance is observed and measured, and outputs
are recorded. Here, then, is a significant extension of the simulation
technology that provides powerful means of assisting in the design,
3
development, evaluation, and improvement of total systems."
A simulation facility has just been completed at the Naval
Missile Center, Point Mugu, Callfornlao It is a 1-1/2 million dollar
building containing 48,000 square feet of floor space which will be
used primarily for weapon system developmento This facility would be
well suited for an operation simulation effort.
3-37
3.2.3 References; S imula tion
1. Robin, F. A., Pardee, R.S., Scheffler, D. L. , Holland, F.C., A Computer
Driven Simulation Environment for Air Traffic Control Studies; WJCC,
Vol. 24, 1963, p. 437.
2. Alexander, L. To, Man-Machine Simulation as a System Design and Train-
ing Instrument; System Development Corp,, SP-331/000/01 , Sept. 27, 1961.
3« Anon, Simulation, BRT-12, System Development Corp.
4. Geisler, M.A. , and Steger, W. A., The Use of Manned Simulation in the
Design of an Operational Control System; WJCC, p. 51, 1961.
5. Cohen, The Design and Objectives of Laboratory Problem IV, RM-3354-PR,
Rand Corp. , Jan, , 1963.
5. Bekey, George Ac, Optimization of Multi-Parameter Systems by Hybrid Com-
puter Techniques, Part I, Simulation, Feb. 1964, p. 19.
7o Bekey, George A., Optimization of Mul ti -Parameter Systems by Hybrid Com-
puter Techniques, Part II, Simulation, March 1964, p. 21.
8. Grabbe, Ramo-Wooldr idge. Handbook of Automation Computation and Control,
John Wi ley & Sons, Inc. , 1961 .
9. Kepcke, J., Computer Simulation of a Complex Secure Communications Sys-
tem, Eastern Simulation Councils Mtg. 16 July, 1962.
10. Arnold, C. R. , Digital Simulation of a Conformal DIMUS Sonar System,
Phar.e I, AD-265398, 28 Feb. 1961, p. 37.
11. Bishop, Wo A., and Skillman, W. A., Digital Simulation of Pulse Doppler
Track-White-Scan Radar, IRE Internat. Convention Record, Vol. 10, Pt. 4,
p. 94.
12. Hicks, C. L. , Analog Simulation of an Acquisition and Tracking Radar
System with Command Capability, Eastern Simulation Councils, Mtg. 16
July 1962.
13. Daev, Do So, Serdinov, A. I., Tarkhov, Ao Go, Model Simulation of Prob-
lems Bearing Upon the Method of Radiowave Sounding, Izv Akad Nauk SSSR.
Ser. Geotiz, 1963, No. 6, p. 936 or (English trans.) Bull. Acad. Sci.
USSR, GeophySo , Ser. No. 6, p. 573 (June 1963; publ . Oct. 1963).
3-38
14. Hara, Hiroshi H. Special Techniques for Two-Dimens ional Air-to-Air
Missile Simulation, Simulation, May 1964, p. 29.
15. Meissinger, H. Fo , Simulation of Infrared Systems, Simulation, Marcii
1964, p. R-23„
16. Anon, Real-Time Automobi 1 e Ride Simulation, WJCC, Vol. 17, 1960, p. 285.
17. Ashley, J. Robert, On the Analog Simulation of Mechanical Systems with
Stiff Position Limit Stops, Simulatfon, May 1964, p. 2K
18. Katz, Jo Ho, Optimizing Bit-Time Computer Simulation, Commun. ACM 6,
Nov. 1963, p. 679.
19. Smith, William Eo , A Digital Systems Simulator, WJCC, Vol. 11, 1957, p. 031
20. Anon, Simulation of an Information Channel on the IBM 704 Computer, WJCC,
Vol. 15, 1959, p„ 87o
21. Anon, Simulation, BRT-12, System Development Corporation
22. Anon, General Purpose Systems Simulator II, Reference Manual, IBM B20-6346,
23o Anon, RTDHS Primary Site Programning System, Informatics Inc.
3-39
3.2.4 Simulation for Command and Control System Checkout
3.2.4. 1 Introductory
Simulation has become a very popular scientific term. It has been
applied to a wide variety of unrelated activities: the numerical integra-
tion of equations of motion, management and war games, pilot trainers,
sociological and psychological experiments. It seems that the term
"simulation" is used whenever any a c t i v i ty is represented by something else.
Simulation is also applied to the activity of system checkout. The operation
of a system is often initially checked out with test inputs which are not
received from the normal or "real" environment. This mode of operation is
popularly referred to as a simulation mode.
Electronic circuits are commonly checked with signal generators and
oscilloscopes. The signal generator is used to supply an input signal
to the circuit while the oscilloscope displays the way in which the circuit
transforms the signal. In simulation jargon, the signal generator would
be termed a signal s imu lator .
A similar approach is used to check out command and control systems.
This paper describes the simulation checkout of three command and control
systems: the Range Safety System, the Real-Time Data Handling System and
MTDS.
3.2.4.2 Range Safety System
The range safety system of the Pacific Missile Range is a complex of
radars, communication links, computers, command and control devices, etc.
Part of the mission of the Range Safety System is to provide range safety
support during missile and space launches from Pt. Arguel lo or Vandenberg
AFB . The range safety function is controlled from the Range Safety Control
Center at Pt. Arguel lo, California. Here radar tracking data is collected
from many radar sites.
3-40
The data is processed In real-time by an IBM 7090 and displayed in
the Range Safety Control Center, The displayed information is used to
evaluate the performance of missiles. If the missile violates any pre-
determined limits, It may be destroyed.
A set of computer program parameters must be prepared for each launch.
These parameters are the characteristics of the missile, local weather
data, program control parameters, etc. Before each launch, a simulation
is run which tests the parameters and the equipment in the Range Safety
Control Center.
The simulation is controlled by the computer program. When the
computer program Is In the simulation mode, it reads simulated radar data
from magnetic tape instead of reading data from the radar Input buffer. The
simulated data is actual raw radar data which Is recorded from a previous
similar launch. The simulated data is processed by the computer program
In the normal fashion. The program output exercises mast of the equipment
in the Range Safety Control Center, This equipment includes digital-to-
analog converters, plugboard switches, plotting boards, control consoles,
etc.
The simulation checks the operation of the program and terminal
hardware but not the radars or communication links. At present the radars
are checked out manually by the radar operators at each site.
The Range Safety System has been built up and modified over a period
of years. It is a patchwork of many smaller systems. The checkout of
all the component systems is a laborious task which must be performed
for each launch. This procedure is coordinated by voice communication.
If the computer could monitor or control this routine checkout opera-
tion, the operation of the Range Safety System would be more efficient.
At present only a few launches can be made each day because of the time-
consuming preparations.
3-41
The checkout of the Range Safety Center is a relatively simple
procedure because it is under computer controK Routine procedures such
as system checkout should be controlled by computers whenever possible..
3.2. ^o3 The Real-Time Data Handling System
The Real-Time Data Handling System (RTDHS) being implemented at
Pt . Mugu, California, is a mul t i -computer system consisting of peripheral
computers and primary computerso The peripheral computers receive and
process radar data at each radar site and transmit the processed data
to a primary computer. The primary computer processes the data,
presents displays and performs control functions. A typical control
function would be the transmission of aircraft vectoring commands.
The simulation checkout of RTDHS is similar to that of the Range
Safety System. Simulated radar data can be read from magnetic tape and
used to check out the computer program and associated equipment. However,
the RTDHS simulation can be more comprehensive than the Range Safety
System simulation. This can be accomplished by transmitting the simu-
lated radar data to the peripheral computers for processing. After
processing by the peripheral computers, the data can be transmitted
back to the primary computers. In this way, the hardware and programs
at each radar site and the transmission system can be checked out in
addition to the operation of the primary site.
Mul ti -computer systems, such as RTDHS, are readily adaptable to
simulation checkout because of the flexibility of program control at
many places in the system. RTDHS simulation modes can be expanded
simply by modifying the computer programs. For example, each
peripheral computer could read simulated radar data from tape and
transmit the data to the primary computer. The radars could also be
included in the simulation because they can be controlled by the
computer program through digital-to-synchro converters.
3-42
3.2o4c4 MTDS
Simulation is used in similar fashion in MTDS. The MTDS configura-
tion resembles that of RTDHS <, It consists of a central or primary
computer which receives the data from a number of satellite computers.
The centra] computer, a Q-20, is used to support the Tactical Air
Command Center (TACC) which monitors the entire "battle". The Q-20
is used primarily to control the various displays in the TACC. A
satellite computer supports the operations at a Tactical Air Operation
Center (TAOC) o The TAOC s identify, classify and assign weapons to
airborne targets and transmit their actions to the TACC.
The MTDS simulation checks almost the entire system. Targets are
generated by the Q-20 and transmitted to the TAOC ' s where they are
processed in the normal fashion. The TAOC's transmit their results to
the TACC for display and command/control action.
MTDS also employs several other smaller simulations for checking
out system components. The operation of the TAOC's can be checked out
individually without involving other parts of MTDS. This is done by
supplying simulated targets to the TAOC with a target simulator, the
SPS-T2A. The SPS-T2A is a built-in piece of hardware which generates
controllable targets.
The best director of MTDS simulation, Major Barnard, made a comment
on MTDS simulation: "Simulations should be designed so they may be
set up and operated completely by military operations personnel. SDC
prepared film which was used for the MTDS simulation runs and the time
required for the film preparation was too long."
The simulation checkout in MTDS is quite extensive. The simulations
which are used to checkout system components are valuable trouble
shooting and maintenance aids. These component simulations can be used
prior to a total system simulation to avoid using the entire system
to locate a malfunction in one component.
3-43
In conclusion:
1) Simulation is an effective method of checking out a command and
control system.
2) Simulation can be much more comprehensively in a mul t i -computer
system because more equipment can be included in the simulation
3) System components should have built-in simulation capability
so they may be checked out i nd i v idual ly »
3-44
3-2.5 Simulation for Command Control System Development
3.2.5.1 I ntroduct ion
The preceding paper on simulation discussed the use of simulation for
the design of command and control systems. This paper will discuss the use
of simulation in the development of command and control systems.
When does the design phase end and the development phase begin? The
time of departure is not well defined. The development of some system
components may begin before the design phase is finished. However, in order
to form a basis for the organization of the paper, an arbitrary time of
departure will be defined,
A command and control system (or total system) is composed of a number
of subsystems: communication systems, detection systems, weapon systems,
transportation systems, data processing systems, programming systems, etc.
The task of total system design involves integrating various subsystems
into an optimum geometrLoal and functional configuration. Part of this
task is determining the general characteristics of each of the subsystems.
For a detection subsystem, for example, such characteristics could be range,
track capacity, accuracy, watts, size, weight, etc. After the general
characteristics have been determined for each subsystem, the development
of the total system can begin, i.e., the design and development of the
various subsystems. This shall be our defined time of departure.
3.2.5.2 General Considerations
The problems involved in designing and developing subsystems are in
general the same as those involved in designing and developing total
systems, i.e. integrating a large number of components into an optimum
configuration. Consequently, much of the general discussion contained in
the first paper will also apply here. In order to avoid repetition, only
a brief summary of analysis applications is given in this paper.
3-45
3.2.5.2. 1 Analysis
Simulation plays a major role in the analysis phase of system design.
Most of the published material on simulation is classified by ASTIA under
"Research." As this indicates, simulation is an important research tool.
As such, It is used in the analysis phase of system design. One of the first
tasks in the analysis phase is to evaluate a large number of alternative
designs. Many times, simulation is used for this purpose because analytical
methods are too difficult to apply. For example, the design may contain many
non-linear relationships which would eliminate the value of a linear program-
ming solution.
One of the largest applications of simulation is in the field of dynamics
The differential equations or mathematical models of moving vehicles are often
too complex for analytical solution. Sometimes the models contain empirical
tables, such as atmospheric density functions, which must be represented
analytically with series approximations. Consequently, an accurate solution
can only be obtained by numerical integration. Although numerical integration
solutions bear little resemblence to man-machine or "madnlne shop" simulation
techniques, they are popularly referred to as simulations.
3.2.5.2.2 Optimization
Many of the references cited in this paper state that simulation was
used to optimize a system design. System designs which contain only a few
parameters can be optimized by evaluating all possible designs. Suppose a
system design contains only two parameters and each parameter can assume ten
values. All possible cases for a design of this type can be evaluated with
one hundred simulation runs.
Unfortunately, the performance of few system designs are dependent on
only two parameters. The optimization of multi-parameter systems is much
more difficult. The gradient or "hill-climbing" method is generally applied
to mu 1 ti-parameter optimization problems.
3-45
The gradient method begins with an estimate of the set of parameters
which will optimize the design. This set of parameters is gradually adjusted
by small steps until it is no longer possible to optimize the design by
further adjustment of the parameters' values. In other words, the top of
the "hill" is reached and movement in any direction would be downhill.
The inherent power of the gradient method is the technique by which the
parameter adjustment is controlled. The amount by which each parameter is
adjusted varies with every step. However, at each step, the vector sum of
all the adjustments is constant. The amount by which a parameter is adjusted
is proportional tDthe sensitivity of the optimization function to that
parameter. Parameters which affect the optimization function the most are
modified by a greater amount. This method follows the steepest path up the
hill.
c -J
Figure 3-5 is a graphical representation of the gradient method. '
_ ,.
Parameter Adjustment
-^
J"
System
Outputs
System
Simulation
on
Computer
Performance
Criterion
Computation
Criterion
Function
F
Computation
of
Optimum
Parameter
Values
•
•
•
Figure 3-5, Block Diagram of Design Optimization Problem
3«47
The gradient method is frequently implemented on hybrid computers.
The analog portion of the computer is used to simulate the system design.
The digital portion is used to evaluate the performance of the simulated
system and to control the adjustment of parameters. The optimization of
automatic control systems is one of the largest applications of hybrid
computer optimization.
3.2.5.3 Subsystems
The examples presented have shown where simulation has been used in
the development of various subsystems of command and control systems. They
are intended as an indication of simulation applications and not as a
comprehensive treatment.
3.2.5.3.1 Communication Systems
Communication or transmission systems can become quite complex,
especially in a decentralized command and control system. Figure 3-
i 1 lustrates a complex transmission system which could be analyzed with
simulation techniques.
Source
Receiver
Carrier
Coder
Modulator
Demodulator
Decoder
Transmitter
Processing
Center
Figure 3-6, Carrier Transmission System
3-48
Analog simulation has been used at General Electric to analyze a
9
secure communications system design. The simulation was used to evaluate
system feasibility, to determine optimum system parameters and to evaluate
system performance in various signal environments. The simulation results
saved time and expense by eliminating the construction of hardware equipment.
Simulation can be used to analyze the performance of filter methods for
reducing the effect of transmission noise or ECM.
3.2.5.3.2 Detection Systems
Radar, sonar and infrared detection systems have been studied with
simulation techniques. Simulation can be used to study the system
performance as a function of the system errors, for optimization and improve-
ment of system parameters and for analysis of measurement accuracy and track
abi 1 I ty .
Analog simulation has been used at General Electric to improve radar
system design concepts. Potential areas of difficulty were illuminated by
12
the simulation early in the design phase.
An article published in Russia describes how digital simulation is
used as a research tool for studying electromagnetic fields around disturbing
13
objects, I.e., plates, discs, cylinders and spheres.
3.2.5.3.3 Weapon Systems
The complete assortment of simulation tools can be used in the design
and development of weapon systems.
Digital computer simulation can be used for the solutions which require
great accuracy. Digital simulation is used for determining guidance para-
meters of ICBM's and space vehicles. The successes of missile launches
demonstrate the accuracy of these simulations. Unfortunately, accurate
digital simulations require a large amount of computer time. Many times
the calculations must be performed in double precision arithmetic.
3-49
On the other hand, analog simulations require relatively little computer
time but are not highly accurate. Consequently, analog simulation is used
when many solutions are required. For example, analog simulation can be used
for analysis of guidance techniques or the calculation of kill probabilities
of air-to-air missiles.
Man-machine simulation can be used to study the performance of human
components in weapon systems. The ability of pilots to navigate from vectoring
commands is being studied with man-machine simulation at the Naval Missile
Center, Pt. Mugu, California. The pilot sits in a mock-up of a cockpit and
responds to the vectoring commands by manipulating conventional controls. The
position of the aircraft is calculated with analog simulation.
TV missile systems have been analyzed with man-machine simulation in
order to determine the ability of pilots to guide missiles. The pilot is
supplied with a TV picture of a target and stick for guiding the missile.
The missile is simulated with a TV camera that is mounted on a platform
which moves towards a model target. The motion of the platform (the missile)
is controlled by an analog computer which simulates the motion of the missile
in response to the pilot's commands.
Sometimes actual system hardware is studied by simulating the environ-
ment of the hardware component. Infrared seeker components have been studied
by supplying a moving target to the seeker through an arrangement of lenses
and mirrors. The motion of the seeker platform (the missile) and the target
are controlled with an analog computer.
3.2.5.3.4 Transportation Systems
Analog and digital simulations are used to study damped spring mass
systems (suspension systems) of transport vehicles. Analog simulation is
used more extensively because it is better suited for the solution of differen-
tial equations. These simulations are valuable for determining the shock
and stresses on delicate components which must be transported: computers,
communications equipment, guidance equipment, etc.
3-50
General Motors has written a simulation language, DYANA, which is used
to simulate complex damped spring mass systems. The input to DYANA is a
description of the physical system, i.e. geometry, spring constants, damping
coefficients, forcing functions, etc. DYANA translates the input into a set
of differential equations which represent the system. A Fortran program is
punched by DYANA which will solve the equations and print out the responses
of system components.
3.2.5.3.5 Data Processing Systems
Simulation is used at the micro and macro level of the development of
data processing systems. One of the largest applications of simulation at the
micro level is for the checkout of logical circuit designs. Since computer
logic is essentially boolean algebra, it can be represented with boolean
expressions. This type of simulation operates at the bit-time level for the
1 8
checkout of logical circuits.
The application of simulation at the micro level is not limited to
computer circuits. A paper presented at the Western Joint Computer Conference
1957, describes how other computer components can be simulated, i.e. drum
IS <
20
19
memory, word structure, information channels, etc. Simulation has also
been used to study error patterns in computer information channels
Simulation at the macro level is used at SDC for the design of data
processing systems. The following excerpt describes a simulation program which
IS used for this purpose.
"A major contribution to the tools available for data processing
system analysis has been developed in the form of a Data Processing
System Simulator (DPSS) . The DPSS is an extremely flexible general
purpose computer program that provides system performance data on a
proposed new design or modification to an existing design prior to
making equipment selections and commitments or performing any signifi-
cant computer program design. The total system design Including the
software and equipment portions can be subjected to a rigorous
analysis and evaluation early in the design process so that key
decisions can be made in the areas of:
3-51
'T) The kind of equipment to be used.
2) The number of each type of equipment.
3) The kind of data processing discipline and
strategy required.
4) The projected performance of the system under
varying loads.
5) The system's maximum capacity.
6) The system's ability to respond as a function of
loading, capacity, and environment."
IBM has written a simulation language, GPSS, which can be used to study
data processing system designs. The following example is taken from the GPSS
22
reference manual.
"EXAMPLE 3 - RAMAINDER DIVISION: A DISK FILE APPLICATION
In a disk file, the disks revolve at the rate of one revolution per
50 milliseconds. Each disk comprises 100 tracks, and each track is
subdivided into five sectors.
Assume that an access arm has been positioned to the desired track
and that the following sequence of operations is to take place:
OPERATION
1. Wait for desired sector
2. Read record
3. Update record
4. Wait for desired sector
5. Write and write-check record
The beginning of operation 5 must follow the beginning of operation 2
by an integral multiple of 50 milliseconds. If the CPU is timeshared,
operation 3 may cause a delay of unknown magnitude.
3.2.5.3.6 Programming Systems
The programming system which controls the real-time processing in a
command and control system is generally under a continual modification. The
modifications result from improvements or expansion of the system. Simula-
tion can be used to check out these modifications. This is accomplished by
simulating input data to the system which will test the program's functions.
Time,
Ms.
Equipment
25+25
Channel
10
1 1
30+5
" + CPU
?
1 1
60
1 1
3-52
The programming system designed by Informatics for the Real-Time Data
Handling System at Pt, Mugu, California, operates in a variety of simulation
23
modes. One mode uses simulated radar data from magnetic tape to check the
operation of the programming system. Another simulation mode is used to
check out the system hardware prior to an operation.
3-53
3.2.6 The Role of Simulation for Test and Evaluation of Navy Command
Control Systems at Pt. Mugu .
3.2.6.1 I ntroduct i on
An example of the role of simulation is the testing and evaluation
of new and complex systems is presented by this description of the simula-
tion facilities which are operated by the Navy at the Naval Missile Center,
Pt, Mugu, California, Facilities similar to those described can be
extremely valuable to electronic system designers. Not only are they of
value in the examination of tactical and attack parameters, such as range
at time of firing, but also similar, less sophisticated facilities, can be
of great value of the system designer in the evaluation of alternative
design concepts during early phases of design.
3.2.6.2 Mission of Naval Missile Center
The purpose of this section is to show the organizational composition,
functions and responsibilities of the Naval Missile Center.
3.2.6.2,1 Organization
The U.S. Naval Missile Center is a tenant activity of the Pacific
Missile Range. Its management control coordination is provided for the
Bureau of Naval Weapons by the Assistant Chief for Research, Development
Test and Evaluation.
The Naval Missile Center is, therefore, under the military command of
the Commander, Pacific Missile Range, and under the Management Control
of the Bureau of Naval Weapons. Under the Commander of the Naval Missile
Center, the major technical branches are i nter- related as shown in Figure
3-7.
3-54
DIRECTOR
MISSILES
COMMANDER
Technical
Director
DIRECTOR
ASTRONAUTICS
Deputy
DIRECTOR
LABORATORIES
DIRECTOR
OPERATIONAL
SUPPORT
Figure 3-7
3.2.6.2.2 Functions and Responsibilities
Tiie mission of the Naval Missile Center is, in accordance with SecNav
Notice 5450: To conduct tests and evaluation of Naval guided missiles,
their components and weapons systems; to provide services and support to
the Pacific Missile Range; to provide supporting services pertaining to
planning, development, evaluation and training in the field of astronautics
and bio-science.
1) Conduct test and evaluation of Naval airborne tactical
data systems and components.
2) Perform Board of Inspection and Survey Trials for
Integrated naval weapons systems in accordance with
Board of Inspection and Survey Directives.
3) Perform research and development for advanced simulation,
instrumentation, environmental test techniques, and
Improved serviceability and reliability characteristics
of missile weapon systems.
To accomplish these task assignments and the total mission, the basic
line functions of NMC are divided into four basic directorates. These
directorates have the following responsibilities:
3-55
1) Pi rector Ml ssi les
The Director of Missiles has the responsibility to provide
technical plans and direction for the test and evaluation
of missile systems subsystems and components. Under this
directorate are the project offices which coordinate the
following projects:
Sparrow I I I
Bui Ipup/Shr I ke
AIDS
PHOENIX
2) Director Astronautics
The director of astronautics has the responsibility of
planning, prosecuting and managing the astronautics
and advanced weapons programs assigned to the Naval
Mi ss i le Center.
3) Director Laboratories
The director of laboratories has the responsibility for
planning, prosecuting and managing the laboratory activities
surrounding the development, test and evaluation of weapon
systems/subsystems and the life science aspects of the
missile and astronautics programs assigned to the command.
4) Director of Operational Support
The director of Operational Support is required to provide
operational support to the Pacific Missile Range (PMR) ,
Naval Missile Center (NMC) , and as directed to the Fleet
and visiting and tenant units in the areas of aircraft
maintenance, target support and photographic services.
3-56
3.2.6.3 The Simulation Laboratory
Construction is being completed on a new simulation and vectoring
facility for the Naval Missile Center. The laboratory will contain
analog computers and other special purpose electronic equipment, and will
study a large class of problems with prime emphasis on simulation testing
of Navy weapons systems.
3.2.6.3.1 Role and Function of the Simulation Laboratory
The use of simulation in the development and test of missile weapon
systems is not new at the Naval Missile Center. Analog computers have been
used for this purpose since 1950,
The original computers were housed in a single room in one of the
large quonset buildings in the old technical area on the beach. Simulation
activities have grown over the years until now all of three temporary frame
buildings and partsof two others are being used for this purpose.
The most general role and function of the simulation laboratory is
to use simulation studies for all those problem areas which can be
effectively studied by this method. The tremendous new physical plant and
equipment being allocated to this effort speaks for its success in the role
of simulation as a tool for test and evaluation.
3.2.6.3.2 Physical Plant and Equipment
The new simulation laboratory (a $1,500,000 structure) is located
at Point Mugu on the beach south of 20th Street. It will rise on a 300
foot front and contain 48,000 square feet of floor space.
There will be 40,000 square feet of laboratory and office space on
the ground floor. A tower will house aircraft mockups on the second floor
and missile system evaluation laboratories on the third floor.
The facility will be used by NMC for simulation of all parts of
weapons systems by electronic analog computers and for vectoring missile-
carrying aircraft into correct positions for launching missiles against
a i rborne targets.
3-57
Simulation activities will account for most of the laboratory areas
in the building. Space has been planned not only for the computers and
special devices to simulate parts of weapon systems, but also for shops
and laboratories.
The shops will be used to maintain and modify the computers, and
the laboratories to design and build other simulation equipment.
The analog computers currently in use and to be moved to the new
facility are of several varieties. The REAC (Reeves Electronic Analog
Computer) has five consoles, each of which contains approximately 60
amplifiers; and two of which have a bank of six coefficient function
generators. The newest of the laboratories' computers Is the Beckman
EASE 2133. This Is a $200,000 class analog computer and has many impor-
tant features including all electronic multiplication, 20 cycle bandwidth,
and considerable capability for presentation of digital Information to the
operator, both dynamic and printed. This computer has 120 amplifiers,
6 electron resolvers, 40 multipliers, 180 potent lameters and 120 trunk
lines for communicating with external devices.
The oldest of the large analog computers Is the Bendix three-
dimensional flight simulator which is approximately eight years old. This
computer has 88 amplifiers.
The PACE computer built by EAI Is used for small problem analysis
such as checking the roll control device on PHOENIX. This analog computer
is classed at 100 amplifiers.
In each of the above computers, the number of amplifiers has been
referenced, thus providing a reasonable Index to the amount of Inherent
computational capability provided by each analog computer. There is also
some current discussion centering around the acquisition of a digital
computer. This will probably come to exist In what Is currently called a
"Hybrid" configuration.
3-58
3.2.5.3.3 Current Applications
Prominent among the simulation projects being carried on now and
to continue in the new building is a cockpit mockup of the F4B (Phantom l|)
airplane. This was designed and built in the present simulation laboratories
to study the problems involved in attacking an enemy airplane when the pilot
of the missile-carrying interceptor never actually sees his target. The
laboratory studies using the initial capability of the F4B cockpit were
initially concerned with two basic problems:
1) How does a ground or shipboard controller, using a long range
search radar, vector the interceptor airplane into a position
where its own airborne radar can "see" the target?
2) How can the airplane be flown close enough to the target to
successfully launch a missile?
The pilot must depend entirely on information obtained from his radar
system to do this. Hence, with this "vectoring" problem to study, the
most important part of the F4B cockpit simulator is the radar display.
Every effort has been made to have the pilot and his radar observer see the
same displays that would appear in a combat situation.
Closely associated with the intercept evaluation is the test and
evaluation program for the Airborne Tactical Data System (ATDS) . This is
a computer-automated fleet-oriented system with similar objectives.
The cockpit simulator requires three large analog computers to
realistically represent:
1) The response of the airplane to the flight controls.
2) The geometry (or geography) of the problem, sometimes
extending over several hundred miles.
3) Simulation of the electronic equipment aboard the airplane
which transforms the raw radar information to meaningful
di splays.
3-59
The AIDS is typical of a complete weapon system which must be
located in a laboratory where it can be studied in a simulated environment.
This system consists of a high-powered search radar and a number of digital
computers which automatically interpret what the radar sees, display the
information and automatically direct a number of fighter aircraft to inter-
cept enemy aircraft.
The equipment is all carried aboard the twin-engine E-2A (Hawkeye)
a I rplane.
A set of operational ATDS radar-computer-display equipments, as found
in the Hawkeye, is Installed and operating at the Naval Missile Center in
laboratory spaces near the analog computers.
The laboratory ATDS is able to "talk to" and automatically exchange
Information with any operational ATDS aircraft while flying in this area.
The laboratory ATDS will be an Important occupant of the new Simulation
and Vectoring Facility.
By locating the laboratory ATDS near the analog computers, many tests
of the automatic detection tracking and reporting functions of the ATDS
computers can be performed without actually having airplanes in the air.
It is possible to know how accurately the ATDS can do Its job without
actually putting an airplane In the air.
3.2.6.4 A Cockpit Simulator
An Intercept simulator was constructed at the Naval Missile Center
to aid In evaluation of the F-4B/SPARR0W III and Airborne Tactical Data
System weapons systems. The simulator combines an analog computer with a
mock-up of an F-4B cockpit and accessory equipment to simulate, In the
laboratory, the flight of a single F-4B fighter from combat air patrol to
breakaway maneuver in the interception of an enemy aircraft.
3-60
Simulation of the intercept flight is achieved by solving, on an
analog computer, math emat ica 1 equations representing the f ighter- target
intercept dynamics, and by duplicating with operating hardware the cockpit
portions of the F-4B airplane. This duplication includes the a i rborne- inter-
cept-radar controls and displays for both the clear and countermeasures
env i ronment.
The cockpit Itself provides simulation only throughthe navigational
instrumentation. No attempt is made to provide such effects as optical
(landscape), thermal or gravitational effects as is common in simulations
used in preparation for lunar landings.
It has been found, however, that for this particular intercept
problem, the adjustment period for pilots to fly smoothly and effectively
without "feel" is a few days and that this lack of "feel" does not affect
the general validity of the system tests.
3.2.6.4.1 General Statement of Test Objectives
By combining an analog computer with a mockup of an F-4B cockpit
and accessory equipment, the intercept simulator duplicates many of the
flight conditions found in Naval airborne intercept tactics. Such tactics,
as used in current fleet defense strategy, deploy early warning (EW)
aircraft around a fleet perimeter, with F-4B interceptors on combat air
patrol (cap) 100 to 150 nautical miles distant. The early warning radars
contact and track approaching aircraft; the information is processed by
a Combat Information Center (CIC) and, if the aircraft is determined to be
hostile, a CIC air controller dispatches one or more of the patrolling
F-4B's to intercept the enemy. Radio communications from the CIC to the
F-4B pilot guide or "vector" him until he can detect the hostile target
with his own airborne intercept (Al) radar. After detection, the target
is automatically tracked by the Al radar until the pilot has launched his
missile and maneuvers away from the collision area.
3-61
Future fleet defense operations are similar in broad outline, but
will involve the Airborne Tactical Data System (AIDS) and the Naval Tactical
Data System (NTDS) . In these systems, discussed in the following sections,
information is processed automatically by digital computers, and once the
interceptor pilot is assigned to a mission, vectoring information is auto-
matically transmitted, received and presented to him by electronic means.
The F-4B intercept simulator consists, physically, of the following
ma j or units:
Electronic Analog Computer
Pi lot' s Cockpi t
Radar Intercept Officer's Cockpit
CIC Station
AN/ASW-13 Digital Data Communications Set
With these interconnected units, the flight, or any portion of the
flight, of a single F-4B fighter can be simulated from CAP position to
breakaway maneuver. Although no actual motion of the cockpits is involved,
the pilot and radar operator "fly" the F-4B within its designed aerodynamic
limits, receive vectoring commands from the CIC, operate the Al radar in
finding and tracking a target, and respond to radar scope displays and
instrument indications duplicating those in actual aircraft. Countermeasures
(cm) effects, such as voice jamming, chaff, decoys and range jamming, can
be included in the simulation. The intercept simulator allows technical
areas of interest, such as vectoring accuracy of the effects of engineering
changes, to be readily investigated; the results are combined with other
ground tests and with flight tests in an overall weapons-system evaluation.
3.2.6.4.2 Simulation of Intercept Problems
The simulation of the intercept problem is achieved by solving, on
the electronic analog computer, mathematical equations representing the fighter-
3-62
Analog Computers
Pilot
Cockpit Mock-up
T
/
/
/
Aerodynamics
Airborne
Radar
Equations
Equations
/
Display
Generators
(Analog
to Video)
Counter
Measures
Combat Information
Center
Manual Vectoring
PPI Scope
O
Fire
Control
Computer
Target
Generator
Air Controller
r~\
Figure 3-8
MANUAL VECTORING SCHEMATIC
3-63
target intercept dynamics, and by duplicating with operating hardware the
cockpit portions of the F-4B airplane, the CIC station and countermeasure
effects. The computer and hardware occupy separate rooms, but are cabled
together and function as a single unit to simulate a typical intercept
situation or problem.
The electronic analog computer (EAC) Is an assembly of electronic and
electro-mechanical units in which DC voltages are used to correspond to
mathematical quantities or variables. Each of the units In the computer
Is capable of performing one or more mathematical operations on the voltages
(and, therefore, on the mathematical quantities) fed into it. By Inter-
connecting the units to perform all the operations called for by any given set
of equations, an electronic scale model of the mathematical equations Is
thereby produced, and the computer can then be operated to give a solution.
The equations are typically those of engineering or physical systems, in
which mathematical operations produce changes In the variables with time;
variables such as position in space, velocity and heading are examples.
In the analog computer, the voltages vary continuously with respect to
time in a corresponding manner.
The equations necessary to simulate the typical Intercept problem
can be divided into four main groups. The equations are Interrelated In
actual use and the quantities obtained are instantaneous.
1) Aerodynamic Equations
These represent the flight characteristics of the F-4B
aircraft. Quantities such as its acceleration, turn rate
and climb rate are obtained.
2) Kinematic Equations
These represent the position and attitude of the fighter
and target as viewed from the early warning (EW) reference
point. Quantities such as distances north and east from
the EW station are obtained.
3-64
3) Al Radar Equations
These represent the basic geometry between fighter and target.
Quantities such as the elevation and azimuth angles of the
target with respect to the fighter are obtained.
4) Fire Control Computer Equations
These represent identical equations which are mechanized
in the Airborne Missile Control System (AMCS) of the
F-4B, and which produce visual indication on the radar-
scopes of favorable conditions for firing a missile.
Quantities such as the maximum error in heading that the
missile can tolerate and the distance, or range, to
missile are obtained.
The quantities -- In the form of voltages -- obtained from the con-
tinuous solution of the above equations are connected to the various units
of operating hardware via cables from a group of external terminals in the
analog computer. Likewise, quantities obtained from the operating hardware
can be entered into the equations via these terminals.
Full dimensioned hooded cockpits of wood and sheet-metal construction
are provided for a pilot and a radar intercept officer (RIO). The pilot's
cockpit is equipped as essentially with a control stick, rudder pedals,
throttle, instrument panel, Al radarscope and a communications set; the RIO
cockpit is equipped with an Al radar control set, Al radarscope, cummuni cat ions
set and an instrument panel. External to the cockpit and supplementing the
Al radar is a rack of electronic circuitry which performs several functions:
1) Converts analog-computer outputs to video for radarscope
displays.
2) Provides realistic A I radar switching sequences and modes
of operation.
3-65
3) Produces the simulation of enemy countermeasures effects
such as chaff drop, angle and range deception, noise
jamming and voice jamming.
The rack is of modular design so that changes can be made readily.
The CIC station is located in a separate room approximately 50 feet
from the cockpits. The station's main simulation equipment consists of a
plan position indicator (PPl) and a communications set. Supplementing the PPI
scope display, it also provides for the simulation of countermeasures such
as chaff drop, range jamming, multiple targets and decoys.
A fully operating prototype of the ATDS weapons system, is being
evaluated in another section of the Computer Division laboratory. The
evaluation plan calls for a tie-in with the F-4B Intercept Simulator.
Anticipating this requirement, the cockpits were prewired for installation
of the tie-in unit, an AN/ASW-13 digital data communications set. Now
installed, it displays vectoring information automatically from inputs
received from the ATDS system. When the simulator is used in ATDS operations,
the CIC station is normally disconnected.
Suppose a hostile aircraft has been detected at a range of 350 nautical
miles at point due north of CIC station. The target has been determined to
be 40,000 feet above mean sea level and to be proceeding south at a speed
of Mach 0.9. An F-4B fighter on CAP has been assigned to intercept the
enemy. The CAP station is angularly removed 40 degrees east of a line
extended northward from fleet center; the F-4B is initially flying at a speed
of Mach 0.9.
The conditions just discussed are initial conditions which must be
specified and set into the simulator before actual operations begin. Each
such condition may be prescribed over a wide range of values, enabling the
simulation of a variety of intercept situations. The target and fighter
appear as blips on the PPI at the CIC station. When the simulator is turned
on, air controller notes movements of the blip and calculates the heading and
speed the F-4B should take; he then radios this information to the pilot
(communications jamming, if present, will interfere). The pilot manipulates
the control stick, rudder pedals and throttle as he would in an actual flight.
3-66
These motions produce changes in the EAC equations and such changes are
instantly reflected in the cockpits as instrument movements and Al radar-
scope displays, and in the CIC as a scope display, hence giving the effect
of continuous flight.
The intercept can be divided into two phases--sea rch and attack. In
the former phase, the pilot continues to be vectored by the air controller,
while the RIO manipulates the radar control and searches for the target on
his radarscope. Upon detecting the target, the RIO acquires lock-on, at
which time the atuomatic tracking mode of the radar is simulated and the
scope display channels are switched to receive fire-control computer inputs;
the attack phase has begun. The pilot now has on the scope a visual indica-
tion of how to maneuver the airplane to a favorable mi ss i le- launch position,
when to fire a missile (the missile itself is not simulated) , and when to
break away from the impact area. At any time during such a fl ight, the
various CM effects can be switched in or out.
Figure 3-9 is a functional block diagram of the simulator.
3.2.6.5 ATDS Test and Evaluation
The modern tactical environment places increasing demands on the
mobility, flexibility and dispersion of the Carrier Task Force. The task
of gathering, transmitting and processing tactical information into decision
making form for the Fleet Commander and his staff has grown proportionately.
The Airborne Tactical Data System (ATDS) has been developed to satisfy this
need for improved data acquisition and cross-tell to permit rapid command
appraisal of the overall tactical situation, as well as rapid solution of a
mass of detailed problems required for the precise control of the elements
of the Task Force.
The ATDS is, therefore, an airborne system which is designed to provide
both intercept control and early warning to the fleet.
Air
Controller
Voice Jam
/'X.
• On
• Off
Radar
Observer
Pilot
Plan Position
Indicator
(PPI Scope)
Al Radar
Control Set
Aircraft
Controls
Cockpit
Instruments
Al Scope
Displays
Early
Warning
Radar
Automatic Tracking
(Lock -on)
Radar Scan
Mode Search
Forces
Reacting
Forces
Aerodynamic
Characteristics
/'>>_
Al Radar
Characteristics
Aircraft
Kinematics
Fire Control
Computer
Intercept
Phase Select
* Attack
Search
Counter-
Measures
Generators
Target
Kinematics
Al Radar
Coordinate
Position
Figure 3-9
FUNCTIONAL BLOCK DIAGRAM OF F-4B INTERCEPT SIMULATOR
I
3-68
3.2.6.5.1 Evolution and Function of AIDS
A basic role of the Naval Missile Center is to conduct engineering
test and evaluation of Navy Weapons System. From this point of view, the
AIDS is considered to be an experimental system, and the purpose of the
current test and evaluation activities at MNC Is to determine the feasibility
of this concept.
The Bureau of Aeronautics started the development program for AIDS in
1955 and evaluation work began at the Naval Missile Center in 1958. The
original AEW aircraft consisted of a radar, PPI scopes and an operator
with a grease pencil and a voice communication link. He would visually
detect and track the "bogies" and report them over the communication net.
As a result of some considerable experience with this technique, system
designers came to believe that this approach to the AEW function would be
inadequate both because of target saturation and because of the excessive
reaction time required to classify the target and assign weapons to the
threat.
The ATDS evolved, therefore, to provide automated processing of many
functions such as; automat ica 1 ly processing the radar data and detecting
the presence of a target, automatically tracking the target and automatically
reporting this target to some surface activity such as the Naval Tactical
Data System (NTDS), automatically vectoring an interceptor to a point where
its own system can take over control of the Intercept.
Thus, the modern ATDS system has evolved. This carrier based system
utilizes the Grumman E-2A (W2F-1) and has an extensive complement of
associated electronic equipments Including: Display Equipment, Communication
and Data Transmission Equipment, Radars, Identification Equipment and Data
Processing Equipment.
3-69
This complex array of electronics is required to implement the wide
class of functions assigned to these airborne picket ships. The required
system command and control functions of the ATDS include:
1) Detection
2) Acqu i s i tion
3) Identification of Target
4) Evaluation of Threat Potential
5) Weapon Assignment
6) Transmission of Control Data to Interceptors
7) Transmission of Tactical Data among the various
Elements of the Fleet
8) Provide Accurate Navigation Computations
These system functions are to be accomplished automatically, semi-automat ica 1 ly ,
and manually as required by the particular mission objectives.
The ATDS command and control functions indicated above are implemented
by the following subsystems and principal components:
1) Detection Subsystems
2) Navigation Subsystem
3) Communication Subsystem
4) Data Processing Subsystem
5) In-Flight Performance Monitoring
3.2.6.5.2 Test and Evaluation Methods
The purpose, again, of these MNC activities is to test and evaluate
the ATDS. To perform this function adequately, an effort was made to acquire
the prime avionics equipment. These equipments were received and installed
in the laboratory.
3-70
To exercise these equipments, a complex of analog computers and other
support devices such as inertial subsystems and the air data computer were
built by MNC so that the system would function in the laboratory as a complete
system. Tests of the ATDS systems were run both with the laboratory set
and with conventional ATDS craft. Of particular interest here, of course,
is the laboratory-based tests. The test series of the laboratory ATDS was
conducted in the following modes:
1) Test runs using simulated inputs.
2) Test runs in the laboratory using live inputs from
radars scaning targets in the sea test range.
3) Combination of live and simulated inputs.
In addition to these test modes, computer programs for the IBM 7090
were written to do computer simulation of some of the computer functions
such as detection, tracking and vectoring. In these cases, the 7090 programs
are written in such a manner as to duplicate, identically, the computations
performed in the computer equipments of the ATDS craft. Using this technique,
one can exercise the logics of the system with the widest possible choices
of circumstances to verify conditions and tests that would rarely happen
i n 1 ive test i ng ,
3.2.6.5,3 Simulated Environments for ATDS
One of the important aspects of the test and evaluation effort is the
series of controlled laboratory tests. These tests runs in the laboratory
facilitate data gathering and recording and, through the use of simulated
inputs, provide a very large data base for subsequent evaluation.
To contrast the simulated inputs with the equipments being exercised
(see Figure 3-l0), these functional subsystems are described:
The Detection Subsystem has three principal components:
1) Search Radar Set
2) Radar Recognition Set (IFF)
3) Computer Detector
IFF Video
Computer Detector
Coordinate Conversion
and
Height Finding
CP-413 /ASA -217
IFF Mode & Code^
IISP -'
^
ASW-
■14
IFF
AN/APX-7
Computer Indicator
to
x,e , Code
Mode
Raw Radar r,9
X V. h
of taroctc
Interceptor
Navigation
Computer
AS\V-13
/
»
o"
1
Radar
AN/APX-96
Air Data
Computer
1 — ►
Antenna
Bearing
Information
Computer
Programmer
Control
CP-558 / ASA -27
o
Doppler
Radar
IC Intercept
Computer
-
o
weapon
Assignment
ATU Automatic
Tracking Unit
Threat
Evaluation
Inertial
Platform
7
Ownship BDHI
MPC
1 '
^f^
Kineplex
ASQ-b-2
Q^^^^^^
Pilot
Link 11 to
NTDS
CO
i
Figure 3- 10
AIDS Prime Avionics Equipment Configuration
3-72
These components are designed to perform the following functions:
1) Detect ion
The Search Radar supplies raw video to the Computer-
Detector and to the AN/ASA-27 Computer- I nd i ca tor Group
(CRT) displays.
Detection probability for weak radar and IFF legitimate
target returns is enhanced by correlating received signals
on a sweep- to- sweep basis, thereby permitting lower thresholds
than would otherwise be possible. Acquisition of false or
meaningless targets by the Computer-Detector is controlled
by sensing the azimuthal and range duration of radar and IFF
returns (high density target areas), thereby moderating the
effects of ECM, sea clutter or other sources of noise that
would otherwise tend to reduce the target handling capacity
of the system.
2) Identification
The Radar Recognition Set transmits and receives IFF data,
compares received IFF data with previously stored data and
transmits "verification" signals, as well as raw IFF video
for display, to the AN/ASA-27 Computer- I ndica tor Group via
the Computer-Detector. IFF data transmitted to the AN/ASA-27
Computer- I nd ica tor Group consist of:
a) An IFF security check bit generated by the Computer-
Detector to indicate a successful bracket decode of
IFF by the Radar Recognition Set.
b) The numerics of IFF replies received by the Radar
Recognition Set.
3-73
c) An IFF validity bit to indicate successful decode
and comparison of IFF with previously stored IFF
code information.
3) Height Finding and Coordinate Conversion
Target height is determined by special processing of
search radar video. Target position data is converted from
polar (R-0) to rectangular (X - Y) coordinates and, together
with target height data, is transmitted to the Computer-
Indicator Group for further processing and display.
The Inertial Navigation System supplies ownship horizontal velocity,
attitude (pitch and roll) and heading (true and magnetic). This information
together with barometric altitude for vertical displacement, a reference
doppler-derived velocity for accelerometer correction and true air speed,
is processed in the Navigation Computer of the Computer- I ndica tor Group to
produce, among other things, ownship present position, ground speed and
ground track angle, range and bearing to destination, wind speed and
direction and platform correction signals.
The Communication Subsystem has two principal aspects:
1) Communications between fleet elements.
2) Command Data link to and from the interceptors.
The multi-purpose Communications System (AN/ASQ.-52 or MPC data link)
provides two-way digital transmission of target data between surface units
and other AEW aircraft. Transmitted target data consists of such items as:
1) Originator's identity and position.
2) 3-D target position and velocity.
3) Target Identifier
4) Target type, threat and engagement status.
5) Track quality and handover status.
3-74
The Digital Data Communications System (AN/USC-2 data link) is used
to transmit guidance to all interceptors and to receive status data from
interceptors capable of reply. Transmitted guidance data consists of such
items as:
1) Controlled Interceptor Identifier.
2) Target slant range and target ground velocity.
3) I nterceptor/ target range and bearing, attack heading
and time-to-go.
4) Command heading, speed and altitude, target altitude and
action to be taken.
Received interceptor status data consists of such items as:
1) True Ai r Speed
2) Altitude
3) Heading
4) Fuel Status
5) Armament Status
The Data Processing Subsystem is a complex of computer equipment
which has, as one of its principals, the Computer Indicator Group.
Target data received from the Computer-Detector is correlated
(associated) with target track data stored in the Computer- Prog rammer to
update existing tracks and to initiate new tracks.
Automatic tracking of maneuvering targets is by linear filters and
unique three-dimensional adaptive gating techniques utilized in the auto-
matic tracking unit, the special purpose digital computer of the Computer-
Programmer. In addition, since both IFF and search radar video are
available for tracking, friendly aircraft are tracked by IFF and beacon
returns for greater positional accuracy as well as greater blip/scan
ratios than are usually attainable from skin-track. To associate discrete
target reports with established friend tracks, the numerics of Mode II and
Mode III IFF returns are compared with the IFF code data stored in the
3-75
Computer- Prog rammer for friendly elements previously entered into the system.
It Is also determined whether tracks are friendly or unknown, and surface
or airborne. Tracks are cancelled after "n" radar misses where "n" is a
function of track status; i.e., whether the track is tentative or established.
Tracks are also updated on the basis of reports via the AN/ASQ.-52 data link.
Operators monitor the automatic detection, acquisition and tracking processes
and supply supplementary position data to the automatic tracking unit as
requ i red.
The Computer- Prog rammer continually extrapolates the position of all
airborne unknown and hostile targets to determine threat potential to a
previously manually-entered defended point, and to assign an appropriate
threat priority index; i.e., ranks targets in order of threat. Upon estab-
lishment of the automatic threat evaluation mode, the target representing
the greatest unassigned threat is made available for automatic weapon
assignment and is also displayed to the operators. Manually-designated
threats automatically receive the highest priority, whether in the manual
or automatic threat evaluation mode.
In this operator selected mode, the greatest unassigned threat is
submitted to the Intercept computer for Interceptor assignment. Stored
data on the available controlled interceptors is then automatically examined,
and, on the basis of aerodynamic capability, fuel status, Al radar/weapon
capability and time-to-go, the Computer- Prog rammer assigns and computes
and transmits intercept instructions to the Interceptor that can best counter
the threat. This assignment process continues until all available intercep-
tors have been paired with threats. Weapon assignments may be accomplished
manually by the operators as an alternative procedure, in which case the
operators pair available interceptors one-by-one with a selected threat and,
based on the appearance of the display, manually assign one of the interceptors
This procedure is continued until all available interceptors have been
assigned.
3-76
Guidance Instructions are automatically and continually computed
for simultaneous control of engaged interceptors. These instructions are
based on an intercept computer program derived from the characteristics of
weapons expected in the operational inventory. In addition, the terminal
approach path is automatically computed on the basis of weapon requirements
and A! radar characteristics to ensure maximum kill probability. Automatic
transmission of guidance instructions to the interceptors is accomplished
via the AN/USC-2 data link as well as automatic receipt of interceptor
status reports from those interceptors capable of replying via AN/USC-2.
Progress of each engagement may be observed on the Control- I nd i cator CRT
di splays.
Reports consisting of positional data, velocity and category, etc.,
on targets selected by the operators for general reporting (or handover
to other AEW aircraft or to surface elements) are automatically organized
by the Computer- Prog rammer and transmitted via the AN/ASQ.-52 data link.
Similarly, the system is capable of receiving target data via the AN/ASQ.-52
from other elements, of correlating such reports with stored target
tracks, and of tracking and displaying such targets to the operators.
Status and order messages (e.g., hand-over) are also automatically received,
processed and answered.
System performance is automatically monitored in flight by pre-
programmed self- check routines in the Computer Programmer. These routines
are performed continually, periodically or on manual initiation. Self-
checking includes the automatic assessment of adequacy of performance as
well as system status, as displayed on the IFPM test set for operator
monitoring and decisions relating to operation in a degraded mode as
required. Test targets are carried in the system (in addition to live
targets) to provide continual verification of system performance. The
IFPM system also provides a practical means of expediting fault isolation
using only the permanently installed aircraft equipment.
3-77
Simulation of input data is of several forms. The input of simulated
radar data is shown in Figure 3-11 and consists basically of range and
azimuth voltages entered into the system at the point where the true aircraft
sensors would pass on this same information. To simulate these inputs,
two characteristics of the sensor data must be closely imitated:
1) Shape of the pulse
2) Time of arrival
The pulse shape is manufactured in either the IFF simulator and the video
simulator. The time of arrival at the Computer Detector is controlled by
the target generator computer. A computer of some capability is required for
this function to produce a correct equivalent of the three radar returns which
are normally received from a single target. The first return is direct and
allows the distance computation, the second two are bounce returns. The
bounce return allows the computation for target height, knowing the time
lapse between returns and height of the E-2A aircraft.
The other simulated inputs are also analog computer derived and provide
for inputs that would normally come from the inertia 1 platform and from the
doppler radar.
The ATDS laboratory set is capable of operating with both live and
simulated interceptors being directed against either live or simulated
targets. To use the cockpit simulator the flight characteristics of the
type of interceptor it is "pretending" to be are programmed into the analog
computers. The cockpit simulator then relates to the ATDS laboratory
equipment set as indicated in Figure 3-12. The cockpit communicates with the
ATDS system through the ASW- 14 and the ASW- 1 3 data link which would normally
be found in an operational fleet interceptor.
Two simulation sources are associated with the Communication Subsystem
and provide for two types of capability:
1) Playback of previously recorded live inputs.
2) Simulation of messages normally generated by other sources;
e.g., NTDS.
Target
Gener-
ator
Analog
Com-
puter
IFF
Simulator
Video
Simulator
/
Range and Azimuth Voltages
IFF
Timing and Synchronization
of Signals is Controlled by
the Analog Computer.
^
Radar /
y\
Range and Azimuth Voltages
Figure 3-11
Radar Data Input Simulation
' / ' ■'
Computer
Detector
3-78
~7~7~~7~7~7~
Computer
Indicator
Shaded Equipments are
Prime Avionics Group
Video Simulator
/
TTTZ
Computer ^.'
Detector / /
Analog
Computer
V77T
Data from
Controls
Computer ,
Indicator /
use -2
ASW-14
ASW-13
F4B
Weapon System
Simulator Cockpit
Figure 3-12
Using the Cockpit Simulator
Shaded Equipment are
Prime Avionics Group
3-79
The laboratory is able to "monitor" any sea test range operation
and record any data items of value to its test series. These sets of
real world data may be played back repeatedly into the system for isolation
of system errors or verification of corrections made to the laboratory model
of the ATDS.
3.2.6.6. Integration Tests with Companion Systems
The ATDS System is normally considered to provide a far- ranging
extension to the fleet-centered NTDS. It is also possible, however, for the
ATDS and MTDS (marine Tactical Data System) to communicate with one another
and exchange information about tracking and other target reports. In this
case, the ATDS is said to provide a seaward extension of the MTDS.
3.2.6.6.1 General Integration Effort
The Naval Missile Center has been made supervisor for conduction
ground technical tests of tactical data systems. In this role, joint tests
are made involving interaction with ATDS located at Point Mugu, NTDS located
at Point Loma and MTDS located at Santa Ana. (See Figure 3-14).
The primary integration concern is with conducting compatibility tests
between ATDS; NTDS, and MTDS to Investigate interface in the following areas:
1) Language Basis
2) Language Interpretation
In the first area, the interest is syntactic and centers around the
allowable symbols used by the system and the rules concerning the various
symbol strings of transmission. Second, an effort is made to investigate
the relative interpretations of these symbol strings. Particular emphasis
is placed on investigation of possible sources of intra-system error in such
as:
3-80
1) Track correlation
2) Navigation
3) Track quality measures
4) Target category assignment algorithms
5) Mathematical transforms
Verification of compatibility is made by performing joint tests with
communication in pairs between the three installations. And, of course,
the ultimate objective is to achieve an effectively integrated tactical
data system complex.
3.2.6.6.2 Communications
The target reporting function, that is, the air-to-surface link for
communication with the various tactical data systems, makes use of the
Collins Kineplex ASQ.-52. This unit (known as link II) uses the Kineplex
principle for parallel transfer of data. Its relation with the NTDS and
MTDS is shown in Figure 3-13. This data link provides the basic intra-
system communication.
An example of the usage of this link is in providing the MTDS with
inputs from ATDS. In many cases, the ATDS outputs required are elementary
and can be provided by an ATDS simulator. For example: to send one or two
slowly changing targets to assist the MTDS in program de-bug operations
does not require the ATDS, itself, to be tied up.
In particular, the ATDS/NTDS interface problem is investigated by
tracking common targets and then looking at track correlation and other
error sources.
3-81
To/From MPC
Shaded Equipment are part
of Prime Avionics Group.
V7
Video Tape
Record-Playback
Capability
Kineplex
ASQ-5'2
Digital
Control
TDS
Special Purpose
Digital Simulator
Simulated Digital
Inputs - Messages
Figure 3-13
Communication Subsystem Simulation
A TDS
Prime Avionics Equipment Set
MPC
k
'
Kineplex
ASQ-52
n
f
'
Link 11
\
NTDS
Point Loma
MTDS
Santa Ana
Figure 3-14
3-82
3o3 IMPLEMENTATION METHODOLOGY
3.3cl Genera]
This area of the study investigates some of the aspects of
electronic system implementation and change of greatest interest to
Naval System Plannerso For convenience, the work is separated into
the slightly more abstract subject of "System Change" and the
slightly more concrete subject of "System Implementation.." This is
done with the realization that the two subjects are in practice
i ndistinguishably intertwined.
The effort is divided as follows:
1 ) System Implementation Process
a) System Design
b) System Implementation
c) System Specification and Documentation
d) Naval System Implementation
i) Who are the Naval System Planners?
ii) What is the Naval System Planning Envisioned?
iii) What is the Naval System Design Channel?
iv) When and how should Fleet and USMC operational inputs
be included in planning and design?
v) What are ANTACCS' test cell requirements?
In this report material is presented from areas a) and b)o
2) System Implementation Process
a) Planning For the Evolutionary Introduction of EDP
b) Design Change Control
3-83
c) The Naval Design Change Control Channel
d) Testing Design Changes
n this report material is presented from area a)o
3-84
3.3.2 System Design
3.3.2.1 Introduction
This material is presented as a series of very short sections,
each addressing an important concept in System Design, System Engineering
or System Analysis. There has been no attempt made to be exhaustive
in each section nor al 1 -encompass i ng in the selection of subjects to
be mentioned. Rather, the purpose is to touch upon many of the most
important concepts in System Design, provide a general understanding of
each point, and direct the reader to some little-known but very sub-
stantial references.
The subjects touched upon in this report are:
1) Systems Engineering and System Analysis
2) Engineering as Art
3) Solving the Entire Problem
k) Practice Versus Theory
Subsequent short papers will address the topics of:
1) The System Design Process
2) The Design Process for EDP Systems
3) The System Cost Concept
4) Difficulties in Evaluating Large Systems
5) The Role of Analysis in Design
3.3«2.2 System Engineering and System Analysis
System Engineering is that portion of the engineering art'*' which has
to do with the design, production, installation, analysis and operation
of those accumulations of men, procedures and equipment popularly
known as systems.
'' A discussion of the "art" of engineering will follow in the next
sect ion.
3-85
It is important to fix clearly upon the concept that System
Engineering treats of four main types of problems (design, production,
installation, operation) and in the treatment of portions of these
problems uses certain analytical techniqueSc In this respect it is
no different from any of the older fields of the engineering art such
as Mechanical, Electrical, Civil or Industrial Engineering.
The topical literature of System Engineering at times emphasizes
System Analysis to the extent that it begins to seem a subject of its
own, rather than an extremely valuable tool for use by system engineers..
This emphasis is quite natural since analysis and analytical techniques
may be thought of and taught in a stylized and orderly fashion and are
therefore more easily discussed in many circles.
Analytical techniques are used by engineers in all phases of their
work to evaluate as precisely as possible the complex interactions of
various parts of systems, proposed systems, and changes to systems.
A more thorough discussion of the role of system analysis in design
will be presented in a subsequent paper.,
This particular set of papers will discuss primarily the more
non-numerical aspects of System Designo An over-all view of System
Design, System Production (to include procurement), and System
Installation will be covered in a similar set of papers entitled
"System Implementation".
3.3. 2o3 Engineering as an Art
This is a concept which flies in the face of much of our culture,
nourished as it is by the popular press. That the concept of engineering
as an art is not widely understood nor, upon occasion, even popular,
has no bearing upon its truth. An understanding of the design
process requires an appreciation of this concept.
3-86
It is quite likely that any confusion as to the "scientific"
nature of engineering results from the popular misapprehension as to
the role of mathematics and science in engineering. Much engineering
practice is based upon the use of mathematics, and engineering (as an
art) uses any available means to accomplish its ends, including
scientific procedure or datac This tends to obscure the fact that
engineering is essentially synthetic, that, although engineers use
many mathematical and analytical tools, most of them spend their
professional lives putting things together to make their work.. How
they do this is a matter of touch, style, instinct and trainingc
For the good engineer it is also an art'''o
The artistic requirements of all phases of engineering are high,
but perhaps the highest requirement exists in the design phase. In
this phase the engineer operates almost exclusively with concept,
ideas, and relationships until most of the critical decisions are
made. Only then can he see the tin being bent and the wire pulled.
By the time people can "see" the product, it is usually too late to
rectify mistakes by anything other than patching., Occasionally,
blunders are made which cannot be fixed at alK
3c3<.2.4 Solving the Entire Problem
It is self-evident that any engineering project, design or in-
stallation should solve the entire problem, but we only have to look
around us to see that they do not always do so. The gantry that will
not fit over its missile; the spacecraft that does not send back its
TV picture, and the tactical system that can't be assembled in the
dark, are all examples of a failure to solve the entire problemc
" For an excellent discussion of the philosophy of engineering the
reader should read the collected papers of Professor Hardy Cross.
Engineering and Ivory Towers , Goodpasture, R. C, ed . , McGraw Hill,
New York, 1952.
3-87
In almost every Instance these failures belong to one of the
three following classes:
1) Failure to meet mission requirements
2) Failure to stay within design parameters
3) Failure to provide adequately for human contact with the
system
Any of these failures may result for one of three reasons. In the
first Instance the customer does not know or refuses to tell (as hard
as that may be to believe) adequate information about the mission,
design parameters and human engineering requirements., If the engineer
must guess about data he must use, then the customer must abide by
the results of this guessworks The system engineer cannot design
effectively If he must work at arm's length from the customer- When
the customer needs additional information about his requirements, the
money Invested in investigation and study is well-spent to prevent
having to cope with an inadequate system.
The second reason for failure Is that the customer and the system
engineer do not always have the same Implicit meaning to their vocabulary
"Reliability" means one thing to the system engineer In his laboratory.
It means something else entirely to the electronic technician working
in close quarters behind some rotating machineryo The customer and
his designer must make sure that they have a firm mutual understanding
of their vocabularies. Words like light-weight, flexible, expansible,
reliable, etc. must have understood meanings before the designer can
hope to succeed. The responsibility lies in both directions.
The third reason for failure is poor performance on the part of
the system engineer, and failure of the customer to reject this
poor performance. Unfortunately, not all engineers are equally good,
and not all engineering errors can be found by Inspection. A bridge
may be a thing of beauty and structurally perfect, but if it doesn't
clear the next generation of warships at high tide It Is a failure
as a well-done project. This type of blunder cannot often be found
3-88
by inspection. The prevention of such errors can only come from a
combination of painstaking study, talented designers and an alert
interested customerc
A number of expensive errors in systems may be traced to the lack
of competent operational input early in the design stage by the
eventual user of the systemc There is some danger of this in an
environment where one organization designs and buys, while another
organization uses the end productc As competent as the designer and
buyer may be, he cannot feel like the user does.
The system designer, the procurement agency and the actual
operational user must establish a qualified informal design committee
early in the design effort to ensure that the mission, the design
parameters and the human engineering requirements are approximately
stated and fulfil ledo
3.3»2c5 Practice Versus Theory
In the practice of engineering (particularly design), there is
a constant necessity to integrate various standard practices, fundamenta
theory, "horse-sense", and ingenuity to the end of producing the most
appropriate system, structure or product for an appropriate price in
an appropriate time.
The proper balance, obviously, lies between the two extremes
of all theory or all field practice. We used field men with their
muddy shoes and their test gear. We also require the numerical analysts
But good design is not an "either-or" proposition. We must have our
designs created, not by theoreticians, nor by pragmat i sts, but by men
with a good appreciation of both the theoretical and the practical.
As obvious as this point is, it has been overlooked in the design
and implementation of many military systems. Some systems which are
theoretically acceptable cannot be taken down, moved and reassembled
with any degree of convenience, although they are supposed to be
3-89
mobile. The striking of this proper balance, which includes both
theory and practice, is not as simple as it may seem^'o
Analysis goes hand-in-hand with design, and the process of
analysis must ultimately provide the designer with data of some
practical impact upon the design problem at hando The designer is
not interested in the small differences between various methods, but
in the calculation of a result which will be meaningful in operation
in the fields The theoretical human limit of the number of interceptors
one air controller can handle by voice is of little real interest if
it far exceeds the radio channel capacity of the station he mans, or
if all intercepts in this time period will be controlled by data link.,
The intellectual challenge of developing new tools or theories is
thrilling, but the customer in a design contract is paying for a
workable design.
During the operation of any system temporary conditions will develop
such that components will operate close to or in excess of their theoretica
limitationc This is particularly true of the man-machine interface.
Much of the art in good system design lies in determining where these
overloads can be tolerated and where we must spend the money to
eliminate themo The reverse of this lies in recognizing those field
conditions which reduce theoretically allowable operational loads
upon equipment, operators or communicat ion «
Another facet of system stress is that which deals with how systems
are handled in the field. They are over-heated, over-cooled, dusted,
moistened, bumped, jolted, etc., often far in excess of what the
requirements anticipate. A good designer will allow for as much of
-'' An excellent discussion of this problem is found as Chapter 1 of
Design of Modern Steel Structures, Grintner, L. F.., Macmillan Co.,
New York, 19^1 .
3-90
this treatment as possible and often will reduce the physical load of
a certain part, slip in an extra gasket or an extra spring to protect
the critical part or assembly.
Tactical systems' operating environments probably cannot be
predicted with any high accuracy, and the concientious system designer
will do what he can to anticipate field abuse and abnormal conditions.
Certain theoretical practices must be tempered by a great deal
of judgment when designing field systems. Very large and very small
parts, such as Cannon Plugs are difficult to assemble at night or
with very cold hands. Certain types of patented huts or shelters may
be erected only twice before critical parts fail, though theoretically
(and in the sales brochures) they are satisfactory. These types of
limitations upon the application of design theory to actual practice
must be carefully considered by the system designer.
The extension of these remarks seems clear. Tactical Mobile
Command Control Systems should be judged partially (but critically)
by their susceptibility to being assembled at night in a rainstorm
under blackout conditions. For exercise, one should apply this test
conceptually to MTDS or ARTOC and use hungry, tired men who have
been shot at seriously that same day.
3-91
3o3.3 System iniplementati on
3o3o3c,l Genera]
Implementation Is a v^ry l^-road and ii]-defined subject which deals
with the problems and processes of designing^ producing^ testing^ and
installing sysiemso A general description of the process is applicable
to the implementation of v;eapons systems^ electronic systems^ and EDP
systems. Hov;ever^ the subject matter of the ANTACCS effort in this area
is the im;p iementat ion of EDP systems^ and particularly those for tactical
command -control systemso
Detailed preliminary analysis defined some 150 steps^ decisions^
processes and products involved in the im.pl emen tat ion of command control
EDP systems. The complex linkages and relationships between these steps
made study and analysis quite difflcultc in addition^ any sort of
graphic representation was unwieldy in the extremeo To discuss the area
properly and build concepts correctly these 150 items were abstracted
and combined Into 80 major steps. These concepts are presented only
graphically in this report.
Im.plementation is divided into seven phases according to Air Force
System Command Terminology;, and this terminology is used in this report.
The use of this terminology is not a final choice^ but some substantial
portion of industry and the military is acquainted with its meaning^
Very considerable work remains to be done in the implementation
area to develop and correlate concepts of particular interest to ANTACCSo
It would be a simple thing to adopt AFSC-ESD-Mi tre-SDC terminology en
massej, but that is not the purpose of the implementation area's efforto
Substantial changes will be made to this data; although it does
approximately represent what must occur in the implementation of EDP
systems for command control o
3-92
3.3.3.2 Discussion
Tlie figures which follow show the implementation cycle divided
into the following seven phases:
a) System Definition
b) System Des ign
c) Program Design
d) Program Production
e) Program Test
f) System Test
g) System Operation
The definition of these phases has long been established by custom
and usage; but our only concern here is to follow the general concept of
each phase and its place in the overall scheme of implementation.
An open style of notation has been used to portray the contents
of each phase. This was used for two reasons which deserve mention here:
a) In nearly every phase most activities have inputs for all
activities that follow in time. This makes for too much ink -
that transmits too little meanlngo in every system all parts
are closely Inter-related In many different respects. This
holds true In Implementation of systems.
b) Arrows and lines convey the Impression that the data shown
is accurate or final or that It should be related In the
manner demonstrated. This Is not the case here..
Events progress in time from left to righto Events stacked top
to bottom take place about the same time, although one or more blocks
can move right or left In any phase - and In actuality do so when real
systems are Implementedo
3-93
System Definition Phase (Figure S-IS) :
This activity is concerned with finding out what the problem is
and what the resources are that may be applied to its solution.. This
Is not the same phase as Mro McNarnara's Program Definition Phaseo After
requirements are defined,, a System Manager is appointed and he sees that
the overall system requirements are further defined according to the
various subsystems to be used.
System Design Phase (Fiaure S-IS) :
This phase begins with selection of sourceso It defines
schedules and quality criteria. During this phase^ the Operational
System Description is prepared, evaluated and concurred upon. Changes
to the OSD are reflected by changes to the System Requirements.
Prnnr^m H^c; i gn Pha^e (Fioures 3-17. 3-18):
This phase is shown on two figureso The phase begins with
comprehensive agreement upon computer, hardware and software design
constants and details. Work is commenced on the overhead computer
facility. The program system design is set, and comprehensive plans
are begun for over-all system testingo Whatever work required on program
conventions and standards is done, and the data base is designedo
As the phase continues, the EAM support facility is begun, the
program system design is evaluated, and the collection of data base
information Is begun. At about the half-way point of the phase, EAM
operating procedures are set, and the procedures for processing program
design changes are established.
The planning for system testing has been continuing and now
matures into defined system tests and schedules for their performance.
At about this point, program design activity is initiated for operational,
utility, data base, and system exercise production programs.
Statement
Of The
Problem
Mission
Objectives
Analysis
Resources
Analysis
Economic
Analysis
Operations
Analysis
Determine
System
Req' ts.
Establ ish
System
Manager
Conf igurat ion
and Interface
Specification
Figure 3-15
SYSTEM DEFINITION PHASE
Set Req' ts
Hardware
Systems
Set Req' ts
Software
Systems
Set Req' ts.
Human
Actions
Set Req' ts
System
Training
Set Up
End I tern
Control
Set Req' ts.
Acceptance
Tests
CO
I
-pi.
o
Send Out
RFP/RFQ
Source
Selection
Determi ne
Schedu les
Set Up
Schedu le
Hon i tor
Set Up
Des i gn
Change Comm,
Prepare
OSD
Eva luate
OSD
Determi ne
System
Reqts.
Changes
Concur
on
OSD
Figure 3-16 - System Design Phase
I
CO
en
0.
Begi n
Prel imi nary
Computer
Faci 1 i ty
Instal 1
EAM
Faci 1 i ty
Establ ish
EAM
Proceedures
Set Program
Des i gn
Conventions &
Standards
Establ ish
Computer
Hardware
Software
Design
Establish
Program
System
Design
Evaluate
Program
System
Design
Establ ish
Program
Design
Chanae Proc.
Establ ish
Data
Base
Establ Ish
Plan For
System
Test
Col lect
Data For
Data Base
Establ ish
Program
Design
Chanqe Cone.
Figure 3-17 - Program Design Phase (j)
I
CO
Des ign
Exercise
Programs
Establ ish
System
Tests And
Schedules
Des ign
Operational
Programs
Des ign
Util Ity
Programs
Eval uate
Program
Des igns
Make
Program
Des, Changes
-©
Design
Data Base
Programs
(a)
Fig. 3-18 Program Design Phase (Ij)
3-98
As the program design phase closes, program designs are evaluated
and changed, and data base preparation begins.
Program Production Phase (Figure 3-19):
The purpose of this phase is to actually bring into being the
coded computer programs. During this phase utility programs, exercise
programs, data base programs, facility programs, and operational programs
are all coded. The computer is delivered shortly after the beginning of
this phase and is made available as soon as possible. Standardized
assembly tests are designed in preparation for the next phase, and system
test design is begun.
Program Test Phase (Figure 3-20 ):
In this phase all five families of programs are parameter, assembly
and system testedc Operational system test materials are prepared.
System Test Phase (Figure 3-21 ) :
in this phase the data base is loaded and the exercise generation
program is system tested. The entire operational system (including
procedures and hardware) is tested. Following operational system testing,
the customer (either the purchaser, the user, or both) performs acceptance
tests.
System Operation Phase (Figure 3-22) :
After acceptance testing, the user puts the system "on-line" and
begins to accumulate the experience and data which will enable him to plan
for changes to his system.
Comment:
This approach to EDP system implementation is quite complex (in
its detail) and organizationally has been made quite monolithic. There
has been some question as to its slow reaction time and high cost. Still,
3-99
it has produced large-scale working EDP systems on time for AFSCo While
most of these functions must be accomplished in some manner, this
discussion should not be considered a recommendation for precisely this
approach for all new systems, particularly those for ANTACCS,
o
Computer
Faci 1 i ty
Avai lable
Code
Exercise
Programs
Computer
Del ivered
— ....... . ,i
Code
Operational
Programs
Code
Design
1 _.,., ... 1
Utility
Assembly
Programs
Tests
Code
Faci 1 i ty
Programs
Code
Des ign
Faci 1 i ty
Data Base
Programs
Programs
Establ i sh
Computer
Operat ion
Figure 3-19 - Program Production Phase
I
o
o
Exerci se
and Operational
Programs
Parameter
Test
Uti li ty
Programs
G>-
Assembly
Test
Uti lity
Programs
Load
Master
Uti lity
Tape
Parameter
Test
Faci 1 i ty
Programs
Assembly
Test
Faci 1 i ty
Programs
Load
Master
Faci 1 i ty
Tape
Parameter
Test
Data Base
Programs
Load
Master
Data Base
Tape
Assembly
r
Test
Parameter
Assembly
Data Base
Test
Test
Programs
-0
Produce
Operat iona 1
System Test
Materials
Figure 3.20 - Program Test Phase
CO
I
System Test
Exerci se
Program
©
Load
Data Base
System
Acceptance
Test
Figure 3-21
System Test Phase
CO
I
o
G>-
System
I nstal latlon
& Train! ng
Changes
i n
Technology
Changes
i n
Mi ssion
Operation
Ma i ntenance
Trai ni ng
Changes
i n
Env i ronment
Evolutionary
System
Change
Figure 3-22
System Operation Phase
Analysi s
I
o
Evaluation
3-104
3.3.4 Planning For the Evolutionary Introduction of EDP
3.304.1 General
The general trend of system planning has become one of evolution
in the past few years. This is particularly true for those systems
which lend EDP support of any sort to the commander while he executes
his command tasko This is primarily true since command tasks are so
complex that it is hard to define completely what help the commander
really needs, and it is then often difficult to develop those EDP tools
and facilities which provide that command assistance.
Not only is the evolutionary approach appropriate for the initial
design and installation of systems, it is also most appropriate for the
introduction of improvements to existing systems. Both the SAGE system
which was abruptly revol utionari ly installed, and the NMCSSC which
was very evolutionari ly developed now incorporate changes by an
evolutionary process similar to the one described below.
This section presents a relatively complete description of the
evolutionary process and how any command or headquarters may begin to
plan for the evolutionary introduction of EDP assistance for command
function. The same evolutionary planning may be applied to the entire
system for those systems with large non-EDP sub-systems.
3.3.4.2 Description
The planning of an evolutionary process for introducting EDP
into a command organization is unique. For identifying the process
as evolutionary emphasizes that EDP development will be dominated by
uncertainty. We cannot anticipate sufficiently how the problems will
change, how commanders and their staffs will profit or suffer from
automated assistance, how the organization will be restructured or
gain new tasks, or modify its scope. These are a few of the unknowns.
3-105
On the other hand, when we demand planning, we commit ourselves
to some understanding of the future, to identification of a range of
plausible and implausible goals, and to the need to decide early on
long lead-time items such as facility space, hardware funding and
procurement, and areas for further research- Accordingly, planning for
evolution is a process of attempting to ensure an appropriate capability
for growth without disrupting current capabilities; but also without
foreclosing on future capabilities (either by being too specific too
early or, equally dangerous, by not undertaking some specific activities
early enough.)
Accordingly, an EDP evolutionary plan handles different problems
in different ways. In some cases it establishes an organization for
attacking the problems without anticipating what the specific solutions
will be. In these cases, the key questions are the size and nature of
the supporting organizations, their interrelationships, and the
procedures for applying and evaluating their efforts- In other cases,
the planning process must recognize long lead-time implementation
choiceso Although it attempts to delay, as much as possible, the time
when these decisions are made, excessive delay will impede future progress;
accordingly, the time selected for making these decisions must consider
trade-offs between uncertainty and delay. Finally, the initial plan
must anticipate the continual need for replanning. It can only do this
if it projects assumptions, milestones, and expected measures of
performance. Over time, these assumptions prove valid or invalid,
schedules are bettered or missed, progress is greater or less. A good
plan will suggest when replanning is called for and, possibly, even the
nature of the corrective action.
Probably the most difficult problems which will need to be faced
in the initial EDP planning is the first area discussed above, that is,
the organizational arrangements for evolving EDP. Before discussing
a possible outline for an EDP Plan, it might be useful to mention some
3-106
of the issues that must be considered in organizing for the EDP support
of an operating command. It is more illuminating to do this in the
context of a specific arrangement.
Evolutionary implementation involves a three-stage development
process. In the first stage, short range improvements are made to
current operational capability and to exercising and evaluation
capability. The lead-time from identification of a needed improvement
to its incorporation in current capabilities is less than six months.
(By incorporation in current capabilities we mean that the indicated
improvement has at least reached the stage of development and testing
that it can be run in parallel with current operational capabilities.)
In the second stage, medium range improvements are developed
and evaluated where these improvements are expected to need a three
month to two year lead-time before they become operational. An
"experimental operations" capability and associated experimental exercise
and evaluation capabilities are maintained to stimulate ideas for medium
range improvements and to provide a test-bed for evaluating these
improvements.
In the third stage, an analytic and experimental center is
operated whose concerns and tools are at a rauch more abstract level than
those used in the centers in the first two stages. The outputs of this
third center assist all agencies in planning and analyzing requirements
and designs. Certain major EDP techniques may be shown to be tentatively
feasible and ready for further development and experimentation in the
second stage. Also, a development program in EDP technical tools is
conducted as a part of this stage. The third stage looks as much as
five years into the future and none of its developments would likely be
operational in less than a year (and then only if they were expedited
with highest priority through the second and first stages). In support
of these three stages, EDP functional design, program design and
3-107
implementation activities specify and develop the short and medium range
improvements, and the experimental models.
In planning the allocation of resources to these various
activities, it is essential to remember that this organization is
intended to provide an almost continuous flow of products and data.
For example, the activity "experimental operations" receives EDP programs
and procedures from two sources; by operating on these it rejects some
products, modifies others, passes them on to current operations, and
develops data for evaluation and further design.
If resources are not properly allocated among the various stages
and activities, serious bottlenecks or gaps can occur. For example, if
relatively inadequate resources are provided to experimental operations,
then it will not have the capability to develop and evaluate the medium
range improvements and inputs from analytic operations. Something will
have to give. The rate of absorbing new techniques from analytic
operations may be sharply curtailed so that this latter activity Is
providing only marginal improvements to the system. The analysis
and verification of medium range improvements may not be adequately
performed so that a higher than appropriate flow of unval idated
techniques is passed on to current operations. Finally, such high
standards for validation may be maintained that the flow of products
to current operations becomes very small, and as a result, the entire
developmental effort is providing few operational improvements.
Fortunately, such a multi-stage development process is partially self-
adapting so that a somewhat balanced flow of products and design data
is achieved. A major role of EDP planning is to monitor the flow of
products through these diverse activities and to adjust the allocation
of resources and the interrelationship between the activities so that a
reasonably efficient and appropriate development organization is
achieved.
3-108
Accordingly^ an initial plan for this development organization
would have to consider such questions as:
1) What resources should be allocated to each stage?
2) What relative emphasis should be placed on design and
development versus exercising and evaluation?
3) Can some of the same facilities be used for both current
operations and experimental operations?
4) What types of experience are required to perform each of
the activities: user, user representatives, analyst, data
processing designers, etc? In managing them? In
planning for them? In monitoring them?
5) How can operational needs be made to guide the development
of technical tools? To what extent are these tools
operationally substantive (eog., planning models) versus
general (e.g., executive systems), versus operational
(e.g., artillery fire support systems.)
6) What documents are required to describe plans, needs,
products, evaluations and tools?
Although these questions have been posed with respect to the three
stage development mechanism depicted in the attached figure, they will
also have to be addressed in the EDP Plan. The plan must also consider
these additional (and possibly more difficult questions):
1) How many stages does the user need in the development process?
2) What is the lead-time for the various stages?
3) What is the role of present agencies in the proposed
mechanism?
3-109
4) New documents will have to be designed, and responsibility
for producing these documents assigned. What is the
relationship of present documents such as Technical
Development Plans and Fiscal Year Functional Requirements
to these new documents?
3o3.4.3 Contents of the Plan
The EDP Plan should address the following areas:
1) Goals and phasing objectives for EDP.
2) Organization and activities for EDP Development.
3) Measures for Change, Allocation and Planning.
4) Current and Imminent Progress.
5) Software Development.
6) Hardware Planning and Procurement.
7) Problem Areas.
8) Proposed Activities.
9) Plan Modif icationo
A brief discussion of the contents of each area follows:
1) Goals and Phasing for EDP . To what extent, over time,
will EDP support be required in ANTACCS to serve operations,
intelligence, logistics, communications, gaming, and planning?
To what extent, over time, can the data bases and processing
routines in support of these functions be integrated? What
other developments will be taking place during the coming
five or so years which will have a major effect on the role
of EDP support? What functional needs should guide early
development activities? Given significant alternate long
range configurations, what intermediate milestones would
3-110
would have to be achieved to attain each long range goal?
What critical decision points exist over time in selecting
between alternate configurations?
2) Organization and Activities for EDP Development. How many
stages should be planned, over time, for developing ANTACCS
EDP? What is the relationship between these various stages?
What documents and other products must be generated in
performing each of these functions? What agencies are
responsible for originating, reviewing, coordinating and
approving the various documents?
3) Measures for Change, Allocation and PlanninQc What
quantitative measures can be applied in planning or
reviewing the growth or change of EDP support? What are
present planning factors for supporting resources (including
various types of personnel) needed to achieve the above
measures? What guidelines exist for allocating resources
devoted to current operations, current exercises and
evaluation, analyses of potential improvements, operational
specification of EDP functions, computer program design and
implementation, development of exercise and evaluation support
and tools, maintenance of EDP systems (including minor
modification), and development of utility systems?
4) Current and Imminent Progress. What is the current manning,
experience and history of the various units using tactical
EDP In the Navy? What EDP capabilities are currently
operational? What EDP developments are scheduled for early
operation? What are the current relationships between the
various services using and developing tactical EDP? How do
present accomplishments compare with past plans and why?
3-111
5) Software Development . How much and what research and
development in software tools should be sponsored by the
Navy? How would these research and development activities
be related to non-Navy RS-D in this area? What developments
can be undertaken which are not operationally specific; for
example, executive programs, time sharing systems, query
languages, data base management systems, modeling ideas,
etc? What user or operational guidance is required in
initiating such efforts and in subsequently monitoring
their development? When might significant new developments
be ready for incorporation in experimental or operational
EDP systems? What steps must be undertaken to ensure that
such new capabilities can be introduced into experimental
or operational systems with minimum disruption?
6) Hardware Planning and Procurement o How should the
procurement of improved data processing, display, communica-
tions and input devices be programmed? What constraints
does the normal programming cycle impose on procurement of
these improved capabilities? Should the programming cycle
be somewhat modified to facilitate the timely procurement
of both major and minor hardware improvements? At the time
of initial installation, how much processing capability
should be reserved to facilitate growth over time?
7) Problem Areas . In preparing any plan, the planning process
generally illuminates problem areas or uncertainties which
fall outside the scope of the planning group or which cannot
be resolved during the planning cycle. What are these
areas? What specific issues and alternatives are involved?
How does the plan cope with these problems? (How soon does
it assume they will be resolved? Does it inhibit certain
specific resolutions?) Can the EDP planning activity propose
a means of resolving some of these problems?
3-112
8) Prnooseci Act!\/it!es » In ii£,ht of the above;, what changes
are recommended to presen'c p-ans including changes in
organizational re i at i o.^sh i 05,, procuremient specifications
and schedules;, and iev^. of o_pporting resources?
9) i^'an riodlf icati nn o Kov.' should che iniuiai plan be revised?
3y vjhom? With whal; coordi .^a : ion and concurrence procedures?
how often?
A numiber of these planning quesiiions are within the scope of the
curren.: ANTACCS and MTACCS efforts^ Ocners rem:ain to be answered as
the Navy develops more information abou: ; cs future operations > '^ne
ihreat and the technology* ic is i nteresal ng to note;, and sometimes a
litcle confusing to be faced v!\'c:\ che similarity between the Total
System Algorithm and the EDP Syscemi Ai cor i chnio They are quite similar
in most respects. Subsequent reports w: ] i show in detail how these
two processes are relatedo
3-113
3.4 SPECIFIC METHODOLOGY
This area is concerned with the investigation of a few pressing
design problems, primarily those in which the designer is faced with
making choices among alternatives* Very little effort has been expended
in this area so far. The effort is scheduled to be directed as follows:
1) Storage and Transmission of Data
2) The Assignment of Tasks - Man or Machine
3) Special Purpose vs General Purpose Displays
4) The Organization of Information Processing
5) Quantitative Design Tools
One paper is presented here on Quantitative Design Tools.
3.4.1 Quantitative Design Tools
This section will eventually consist of a few papers, not necessarily
closely related to each other, but each relating to subject matter of
high interest to the System Planner or System Manager. This particular
paper shows one technique which is of value to system planners in evaluating
the capability of various computing central processes at early stages of
design or planning.
3.4.2 The Calculation of Figures of Merit For The Comparison of Digital
Computers
3.4.2.1 Abstract and Summary
This paper discusses the theory, construction and application of the
Figure of Merit technique for the evaluation of contemporary computer
system's central computers and high speed memories. Four currently
available methods are presented and analyzed (Class Method, Information
Channel Capacity Method, Efficiency Index, and Babbage Method). A new
method is presented (Highland Method) which avoids many of the short-
comings of previously used Figure of Merit Methods.
3.4.2.2 Introductory
This paper discusses several approaches to determining arbitrary
numerical measures for comparing the "computing capability" of electronic
3-114
digital computers. Measures of this nature are often called "Figures of
Merifa The measures discussed here, and others like them, consider
only the "main frame" and high speed memory capability of the computer
being examinedo That is, they consider only the size of high-speed
memory, the speed with which data is transferred into the computer from
memory, and the speed of computation.
Since one of the crucial limitations of modern data processing
equipment is often input-output capability, these "Figures of Merit"
approaches clearly leave much to be desired. However, we must bear in
mind that normally the purpose of computer installations is not to perform
input-output functions but to manipulate data. Regardless of I/O (input-
output) limitations, this work is done by the central computer, and
figures of merit have real value in the comparison of central computer
capability without regard to type of computer or the application for which
the computer is used.
To complete any worthwhile analysis, considerations such as instruction
repertoire, I/O capability, amount and type of low speed storage, mean
time between failures, mean time to repair, etc. must be studied analyti-
cally. Nevertheless, figures of merit offer substantial advantage to the
system analyst who understands their rationale and limitations, and who
confines their use to "rough-cut" first approximations."
3«4.2.3 Rationale
There are two distinct general approaches to measuring the capabilities
of computing machinery. Only one of these (the figure of merit) is
discussed in this paper. To understand this one technique fully it is
necessary to understand the other (the "bench-mark" technique) to a
1 imi ted degree.
An unpublished paper by Mr. Ronald W. Rector is acknowledged. In this
paper (Measuring the "Capability" of Computing Equipment) he cites a
number of figure of merit techniques. A part of this paper draws upon
these.
3-115
1) The Bench Mark Technique
This approach to measuring computer capability is problem
oriented. That is, machines are evaluated on their ability
to perform certain problems or selected parts of the total
task proposed. These problems may be entire real problems,
parts of real problems or synthetic problems made to resemble
real problems closely. This technique is called the "bench-
mark" method since it compares machines by examining their
differential capability (normally speed) to perform the same
"benchmark" problem.
The bench-mark technique (if carefully executed) can be quite
accurate, but it is very costly in talent and time, and requires
an accurate and precise definition of the total task to be
performed. In addition, any bench mark problem which is not
the complete task ultimately to be demanded of the computer
takes on certain aspects of simulation and is subject to many
of the limitations of simulation.
2) The Figure of Merit
This approach attempts to evaluate the capability of an in-
dividual machine without regard to how that capability will be
used. This is much the same thing as a power station being
given a kilowatt rating without regard to how much electricity
is used or how it is used. At first, this may seem a little
foolish since the only reasonable purpose of computers is to
solve real problems. However, system planners find it very
useful to be able to think of and measure main frame and memory
capability in the abstract. Figures of merit permit them to do
this.
Figures of merit may be used to provide prel iminary answers to a number
of problems without the need to prepare a bench mark analysis. Among
these problems are questions such as:
3-116
1) I am now processing data at rate R« My work load will Increase
to about 7R. What various machines should I consider acquiring?
2) My old machine needs to be replaced. What will I have to pay
for a new machine, and how much capability could I have left
for expansion? This Is really a new statement of question H^\ •
3) Company A charges $5,000 per month for machine 1. Company B
charges $7,500 for machine 2. Is the difference worthwhile in
terms of data processing?
4) The new system I am planning should have the computing load of
about half that of System X, which uses a CDC 6600 at about full
capacity. Allowing for 20% expansion what machines should I
think of for my system? I plan to split the computing load
among four computers. A, B, C, and D.
B = "2" 3nd C = -j-
These and other important questions of a preliminary planning and design
nature can be answered by using some figure of merit technique.
The entire figure of merit approach is based upon the premise that
"more" is "better". The question "Is 10% more also 10% better?" will be
discussed later. The more fundamental question "More what?" is answered
(depending upon what figure of merit we consider) by "more internal speed",
"more high speed memory" or some combination of both. How these qualities
are juggled or combined differs from case to case and Is discussed by
individual case.
In general, we can say that more speed is better in direct proportion
to the Increase. That Is, a four-fold Increase in speed Is four times
"better", and a six-fold Increase is six times "better". Another way of
looking at this Is, a machine which can do work In four hours that was
previously done In eight Is twice as beneficial to the user. This is
particularly true of machines used "on-line".
Considering the usefulness of high speed memory to a user, we can
say that more is better, but not In direct proportion to the increase.
3-117
That is, to go from a size of 500,000 bits to 1,000,000 bits is more
beneficial to the user than to go from 1,000,000 bits to 2,000,000 bits -
even though the increase is by the same factor.
There is however some difference in opinion as to how much the worth
of memory changes as size of memory grows larger. The manner in which the
incremental utility of larger memories decreases is generally felt to be
logarithmic (or some function so close to logarithmic that the difference
is not worth worrying about). Remember we are searching for some numerical
way to express professional opinion, so accuracy is greatly to be preferred
to precision. Accuracy is faithfulness of conceptual replication, while
precision refers to the degree of refinement of the measurement. It is
easy to have one without the other, but precision without accuracy is
misleading, at best, while accuracy without precision is often very useful.
For some applications, perhaps one such as message switching, memory
requirements may be thought of as absolute. That is, the high-speed
memory must be big enough to do the job - but size increments beyond that
point are of little use. For these applications, and those where time
constraints are severe, more attention should be paid to the efficiency
of the computation process than is normally done.
With this introduction to the rationale of figure of merit, we may
proceed to the technical discussion of several types of figures or merit,
their applications and shortcomings.
3.4.2.4 The "Classic Method"
Rector has applied the name to this method, and while it may not
be "classic" in the most pristine sense of the word, the method has been
applied in much of the literature. The calculation is a simple one:
Class Figure of Merit (CFM) = log,_ — =-
Where M = High speed memory capacity in bits
and T = Access time is seconds
3-118
Various forms of memory arrangement must be converted to give a
total reading in bits. Sign bits and parity bits should not be included
Access time is tine time required to fetch a word (or character or
set of characters) from memory. In destruct ive- readout memory machines
the data cannot be operated upon until that small portion of memory is
restored with the data just read out destructively. This takes one
more memory access time. The two times together are called a memory
cycle. Most data are given in cycle time and must be divided by two.
However, in non-destructive memory machines operations begin immediately
after access time.
Since most tabular data presents the time in microseconds (The
Adams Chart, for instance) it is most convenient to use, and subsequent
calculations in this paper will use, microseconds. Since there is no
standard, we can use what we wish, but microseconds are more widely
used and more convenient.
By using this method, we can calculate the CFM for many storage
and access cevices, not just computers alone. Some values calculated
in this manner are shown in Table 3-1.
TABLE 3-1
CLASSIC FIGURE
; OF MERIT
Max. Wds.
BIts/Wd
Total Bits
Storage
Cycle Time
(In musecs)
Cycle Time
2
^'ts Log
Access
Bits
10 Access
CDC 6600
262 K
60
15,720,000
0.7
0.35
44,910,000
7.6523
IBM 7030
262K
64
16,768,000
2.2
1.10
15,240,000
7. 1829
Hughes H-330
181K
48
6,288,000
1.8
0.90
6,969,000
6.8432
PhMco 212
65K
48
3,120,000
1.8
0.75
4,160,000
6.6191
RCA 601
32K
56
1,792,000
1.5
0.75
2,389,000
6.3783
Unlvac 1107
65K
36
3,340,000
4.0
2.00
1,170,000
6.0682
SDS 9300
32 K
24
768,000
1.75
0.87
882,800
5.9459
CDC G-20
32K
32
1,024,000
6
3.00
341,300
5.5332
Packard 8.440
23K
24
672,000
5
2.50
269,900
5.4313
CDC 160A
32K
12
384,000
6.4
3.20
120,000
5.0792
SDS 910
16K
24
384,000
8
4.00
96,000
4.9823
Basic Data From Adams (Nov., 1963)
CO
I
iO
3-120
Several points must be completely understood by the system planner
contemplating the use of measures such as this one. These are:
1) The logarithmic nature of the CFM number.
2) The equal treatment of memory and speed increases.
3) The implicit relationship of computation speed and access time.
The CFM is, by definition, the logarithm of a decimal number. Its
being logarithmic has several implications for a user.
The human mind apparently thini<s in linear terms as a normal course
of events. Even when presented with a table and the certain knowledge
that the CFM is a logarithm, it somehow seems more real to think of terms
varying from 100,000 to 45,000,000 than from 4.9 to 7.6. Our world of
experience is linear, and dealing with logarithms can be quite illusory
for those not on guard.
Therefore, when we look at Table 3-1 we may note casually that the
910 is 4.9+ and the 6600 is 7.6+. This would mean to many persons that
two 910's are a little better than one 6600. Of course this is not
true, and the error comes from treating logarithms as decimal numbers.
In reality, the table tells us that the capability of the 6600 is three
decimal places greather than the capability of the 910. And that says
that the 6600 is between 100 and 1,000 times as powerful as a 910.
This is useful information, but it cannot be said that it is
intuitively obvious, as good as the 6600 is. it would have to be worked
out very carefully and for a particular example. We find, then, that
direct comparisons between the very high and very low ratings on the
scale may be open to some question. It is also open to question as to
how meaningful this 1,000 to 1 ratio could be even if it were quite accurate.
The illusory nature of logarithms and the seeming abnormal compression
of the scale should be looked at again. This time look at three computers
clumped at the center:
Hughes 330 CFM = 6.8432
RCA 601 CFM = 6.3783
Univac 1107 CFM = 6.0682
3-121
These machines appear to be very close together in capability,
particularly since they have the same first digit in their CFM. One
might imagine that they are indistinguishably close. By reference to
column A we see that the quotients prior to the taking of the logarithm
lie in the relationship of 6.9: 2.4: 1.2. This is a considerable
difference, indeed, and it is in adjacent areas of this long table that
comparisons of CFM's have a great deal of usefulness and reasonable
credi bi 1 i ty.
We have just backed into three fundamentals of logarithmic tables
which must be thoroughly understood by any system planner who uses the
CFM technique.
1) Logarithmic representations must be used to place extremely
large numbers and very small ones in the same table conveniently,
and to allow these numbers to be manipulated pleasantly.
2) The use of logarithms obscures the true linear relationships of
many types of data, and can simulate logical errors by all but
the most cautious users of these types of tabular datao
3) Arithmetic operations must be performed upon the Antilog of the
CFM not the CFM itself, that is, the quotient before the log,-,
is obtained.
Using the data in Table 3-1 we wi 1 1 solve problem 4 in Section
3.4o2.3« This will crystallize the points discussed so far.
The proposed system will have a load of about one half of System X
which uses a CDC 6600 to about full capacity. Allow for 20% expansion.
A A
Use four machines A, B, C, and C, with B = -^ and C = -r^ We will
confine ourselves to machines from Table 3-1.
CDC 6600 CFM = 7.6523 (1)
Antilog^Q 7.6523 = 44,910,000 (2)
44,910,000 , ,
— ^ — ^ = 22,455,000 (3)
120% X 22,455,000 = 26,946,000 (4)
3-122
We intend to split the load derived in (4) among four machines.
The load must be allocated 6/13 to A, 3/13 to B, 2/13 to C and 2/13 to
C.
26,946,000 ^ 2,072,769 (5)
13
A = 6 X 2,027,769 = 12,166,614 (6)
B = 3 X 2,027,769 = 6,083,307 (7)
C = 2 X 2,027,769 = 4,055,538 (8)
logjQ 12,166,614 = 7.0853 = CFM^ (9)
log^Q 6,083,307 = 6.7841 = CFMg (lO)
log^Q 4,055,538 = 6.6580 = CFM (11)
From (9) we see that a smaller than maximum size 7030 will do
well for machine A. From (10) we see that an H-330 is close to exactly
right for machine B, and from (11) we see that the 212 should be used
for machine C.
The outstanding shortcoming of the Classic Figure of Merit is that
it treats increments in storage as being equally beneficial.
Let us state the CFM equation again:
PPj^ _ , (High speed storage in Bits)
•^ (Access Time in Microsecs.)
The logarithm,^ does not apply to either the numerator or the
denominator, but to the quotient, and therefore treats increases in
speed and increases in memory as equally beneficial. For speed this is
desirable. For memory size this is not really acceptable.
The worth of machines is often estimated by specialists to look
something like
M •*. _ ^09 (high speed storage in bits)
access time in microseconds
3-123
This expression satisfies much of the discussion here and some-
thing like it will be treated later.
In the Classic Figure of Merit and in some others, the only computer
speed considered is cycle or access time. In destructive readout machines,
cycle time equals two access times. Most instructions also require
integral numbers of access times for their execution. This is because
internal speeds are governed by a clock (in synchronous machines) and
hence by how fast that clock will permit instructions to be executed.
Normally, the fastest tasks of logical testing or shifting control
unconditionally will occupy one access time, and more complex instructions
more integral units of access time. Thus, a reasonable approximation of
the internal processing speed may be had by looking at access time.
However, for a really accurate estimate of the internal computational
speed of any machine, reference must be made to instruction time. This
is treated in a subsequent section.
In asynchronous machines, front parts of each instruction may be
thought of as overlapping with the final parts of preceding instructions,
and therefore access time is not as rel iable a measure of computation
speed. Still, computation is wedded to the speed with which numbers can
be shifted into and out of memory, and access time is a reasonable
indicator of that speed.
When these techniques are used with non-destructive readout machines,
extreme care must be taken to use access time for non-destructive machines
and cycle time for destructive machines. This is because in non-destructive
machines computation can begin as soon as the number is brought in, while
in destructive machines one additional access time is required to restore
the number to its original memory location.
In figure of merit computations, considerations other than those of
the main frame, memory and some approximation of computation speed are
entirely ignored. The capabilities of input/output peripheral equipment
for each system must be studied in detail according to the requirements
3-124
of each system, and they are not amenable to approximation before the
requirements of a system are reasonably well knowno It must be remembered
that some relatively slower machines have fine input/output and peripheral
equipment and, thus, more than make up for their so-called "speed
deficiencies".
3.4.2.5 Information Channel Capacity
Data processing machines that are used primarily for switching
purposes and have memories which meet the absolute minimum required by
the problem, may be compared by the use of a slightly more involed tech-
nique which treats only the internal speed of the computer."
Channel Capacity or C = .Q.
N + T
P
Where L = Word length in bits
N = Number of bits required for the execution of an operation
P = Clock rate in bits per second
T = Average wait time
Q. = Number of simultaneous operations performed.
This approach does yield a good measure for the internal effectiveness
of a computer used solely as an information switch. Its shortcoming is
primarily that, since the approach does not consider memory requirements
as other than absolute, the approach has little general application.
This method also has the disadvantage of considering word length
(longer = better) without considering memory size. The result of this
is two-fold. First, machines with long words come out better than machines
with short words - even if they have the same number of bits in memory,
which is hardly reasonable. Second, it is quite possible for a machine
with the longer word to be less efficient (even given an equal -sized
memory) than a short worded machine, for the following reasons.
* This technique was developed by Amelco, Inc. In a study performed for
Douglas Aircraft as a part of the Army/Navy Instrumentation Program.
Data Processing. AN IP Research. June 1961, Amelco, Inc.
3-125
Most command control processing^ indeed much business processing,
consists of setting and testing items (parts of words) not of making
2 3 4
arithmetic computations using full words. ' ' To do this, a word
with many bits must be shifted or cycled a larger average number of bit
positions than a word with fewer bits. This takes more time. There
are machines having special logical circuitry which allows the testing
and setting of a few bits without manipulating the entire word. In
other than those machines, it is misleading to say "the longer the word,
the better". Often this may be completely incorrect. This argument
assumes the same number of bits in memory, of course.
However, the reason for including this number (L) in the computation
here is: The more bits in the word, the more data can be transferred in
from memory in parallel, and this is an advantage - though somewhat
diluted sometimes by an increase in shifting time.
As with other figures of merit, this one does not evaluate input/
output or peripheral equipment. It is included here primarily to show
a good method for evaluating internal timing.
3.4.2.6 Efficiency Index
The general concept of indices of efficiency is that they measure
the ability of the device examined to produce output equal to the input
provided. A steam engine's efficiency is the ratio of the BTU per hour
output to the BTU per hour input in fuel.
When we compute the "efficiency index" of digital computers, dollar
cost is used as input in the place of BTU/hr input, and the efficiency
measure is supposed to show how much "computational ability" per dollar
cost is delivered by various machines.
One of the many possible manners of computing an index such as this
Is shown below.
Efficiency (E) =
t Ca
Where n = Number of bits per word
t = Add time + 0.01 Multiply time
Ca = Cost of arithmetic and control units
3-126
This measure has several shortcomings. Nearly any measure using
the same terms will have the same disabilities, regardless of how the
terms are accumulated ari thmetical ly.
1) Using the word length alone in the numerator has the same weak-
nesses it had in Channel Capacity measurement.
2) Using cost in the computation of the index itself has three
serious disadvantages
a) It is very difficult to obtain the bare cost of the arith-
metic unit and of the control unit by themselves for a
large array of computers. Granted that it can be done for
any particular computer at will - it is still a formidable
task for the 75 odd computers now available in the U.S.
The G.S.A. electronic supply catalog will have the prices
of the pieces, but customer engineers will have to be
questioned to make sure the correct set of prices is added
up to produce the total cost.
b) The total cost of the various systems is not any constant
function of the arithmetic and control unit. Some computers
have low priced units, others high, and any system must al 1
be bought and installed to obtain whatever efficiency is
inherent in the two units discussed here. It is only the
whole cost of the whole system that is of any importance to
us.
c) Regardless of what cost is used, it is subject to considerable
fluctuation, irrespective of what is published by G.S.A.
This is true since costs are not physical constants of the
machine itself, but are derived by management fiat. By
using rather vague and fluctuating data in the computation,
particularly in multiplication or division, the entire result
is open to the most serious question. Of course, prices
should be considered, but they should be considered separately
from the physical constants of the machine itself.
3-127
3) The most serious consideration in this type of measurement is
the use of
t = Add time + 0.01 Multiply time
Naturally, internal computational speed should be considered in
evaluating any computer. The Classic Figure of Merit does this
indirectly as stated earlier. In this instance, the construction
of the factor t implicitly states that the programs, yet to be
designed and coded, will call for two times access time instruc-
tions (like add) 100 times as often as they will call for 8, 10,
12 or more times access time instructions (such as multiply and
divide). We must not interpret the construction of "t" to mean
that add and multiply themselves will be most popularly used or
will occur with this relative frequency, only that instructions
requiring that number of access times will occur with that freq-
uency. The consideration is this. By constructing "t" in this
way we are, in effect, simulating (or guessing at) the future
use of the computer. If we are close to correct in our guess,
our answers will be very good indeed (barring other flaws in the
computation of these indices). If we are not close to correct,
our answer will be terrible.
It is desirable, however, to get a better reading of internal
computational speed than is done indirectly by the CFM and this
is a very reasonable way to do so. Analysts using this technique
should be aware of its possible shortcomings. That there is some
possibility of error should not prevent the consideration of the
technique.
4) This figure of merit cannot evaluate the efficiency of the entire
computational system since it cannot estimate the input/output
and peripheral equipment accurately (indeed, at all) before the
system is planned. This shortcoming is not peculiar to the
efficiency index alone, but is shared by all figures of merit.
3-128
3.4.2.7 Babbages
C. J. Shaw of SDC has developed, but not documented, a figure of
merit which avoids many of the shortcomings of those discussed previously.
The numerical answer is in terms of "Babbages", a unit of measure he
originated.
The Babbage rating of a computer is obtained by using the following
equation:
B = L log^ M
T
Where: L = Length of word (in bits) transferred to/from
memory during the access time, T
M = Total number of bits in high speed memory
T = Access time in microseconds for transferring
in L bi ts in paral lei
The introduction of the term L in the numerator as a multiplier
gives a much higher rating to those machines which transfer more bits
per access time. This does not mean that, all other things being equal,
longer words mean better computers. It means simply that the more bits
that are transferred in at each access, then the more information reaches
the computer each accesss. In this respect more is better. As was stated
earlier, there is a possible shortcoming here. Machines with proportionally
longer words consume more time cycling and shifting data into the correct
position (once it is transferred in) if they do not have some character
and/or partial word logic, as well as full word logic. The consideration
of this term, then, while highly desirable, is capable of producing some
error if the analyst does not guard against it.
The log„ M term in the numerator states that each successive bit of
storage added to memory is 1/2 the benefit to the user of the immediately
previous bit of storage. This is probably too severe a judgment upon the
marginal value of increments of storage. In most discussions with pro-
grammers and systems analysts, the author has found that the feeling is:
3-129
"Each bit is almost as valuable as the preceding bit. Almost - but not
quite", in all fairness to Shaw, he has admitted that incremental bits
of memory were probably more valuable than 0.5 of the preceding bit, but
that he chose log^ for ease of calculation. There is a mathematical way
around the difficulty of using logarithms to other than the base 1 or 2.
This will be shown in detail later.
There is, however, one real difficulty in the construction of Shaw's
"Babbage". It is an obscure mathematical shortcoming, but one which has
a tremendous effect upon the resultant rating. In the Babbage computation,
the principle is applied inadvertently and is, therefore, a severe short-
coming. This will now be explained.
When the logarithm of a number is multiplied by another number, the
product is the logarithm of the original number, but to a new base.
What this new base is is determined by the number used as the multiplier.
A different number - a different base. The equation governing this rela-
tionship is:
log^y = —J— • 1og,oY (12)
log,o X
This means that we can handily find the logarithm of any number to
any base we desire, given the presence of a table of common logarithms
(log,^)« But it also means that in the Babbage computation the logarithmic
base used to evaluate the size of memory varies inversely as the size of
the word transferred from memory during the access time.
Stated another way, the error says that as the number of bits trans-
ferred from memory gets larger, the more valuable to the user is each
succeeding bit of memory. How valuable is dependent upon what size the
word is; but here are three examples:
The percentage value to the user of each
If thg multiplier is; new bit in terms of the precg<Jing t?its is
6.8 71%
12.6 83%
24.1 90%
3-130
Now it is very likely that each succeeding bit is something from
0.7 to 0.9 as valuable as the preceding bit, as discussed before.
However, it is poor technique to have this value function fluctuate
between computers - depending upon something else entirely. There is
a method to consider word length transferred without encountering this
difficulty, which is discussed later.
An interesting point is that since the log of the numerator is
operated on arithmetically by the formula, the resultant Babbage reading
can be manipulated arithmetically without the logarithmic difficulties
mentioned in the discussion of the CFM.
The Babbage Method goes far toward providing a very useful measure-
ment. It has produced reasonable comparisons when the result was tempered
by some professional judgment. It is clearly the best Figure of Merit
method developed to date. It is worthwhile, however, to examine one
more attempt to provide a Figure of Merit measurement.
3.4.2.8 The Highland Method
The Highland Method of computing figures of merit has been developed
by Ee K. Campbell over the past two years. It represents an attempt to
produce a Figure of Merit method which obviates the internal logical and
mathematical difficulties which appear in those approaches mentioned pre-
viously. This method was developed in an intermittent and evolutionary
fashion. It does not suffer from most of the logical and mathematical
difficulties of other techniques, but is still subject to the inherent
limitations of Figure of Merit.
A .
B
Where: K = Conversion Constant (see below)
M = Total Bits in High Speed Memory
A = Add Time (in microseconds)
T = Memory Access Time (in microseconds)
B = Bits Transferred in parallel during one access time
3-131
K is the constant required to change the log.Q M to the log of M
to another base depending upon what value is selected for K. Table 3-2
which follows, shows some values to use for K, depending upon what
value is selected for the marginal utility of additional memory.
TABLE 3-2 VALUES OF THE MULTIPLIER 'K'
Incremental Va
Additional Bits
lue
9f
of
Memory
M*
Value of
jltiplier "K"
0.40
2.5
0.50
3.3
0.71
6.8
0.77
8,7
0.83
12.6
0.90
24.1
The use of K allows the analyst to adjust the evaluation to reflect his
professional judgment as to the incremental value of memory for the
application at hand. It is reasonable to believe that for most applica-
tions the value of K is somewhere in the vicinity of 0.7 to 0.9, though
for some it could be much higher (or lower). The method of computing
new values for K is as follows:
'°9x ' = ; -, • '°9l0 Y
The incremental value is .
Therefore: if the incremental value of bits added to memory
is to be 0.4,
Then,
0.40
2.5
3-132
and, from the first equation,
1
'-^2.5' =
logjQ 2.5
log 2.5 = 0.39794
K = ^ = ] = 2.5
log,Q 2.5 0.39794
M is the total number of data bits in memory. That is, the total number
of bits excluding sign and parity bits. Log._ is used since tables of
this function are easily obtained, and multiplier K changes log,-, to
whatever base we wish to use.
A is the add time of the machine. It is necessary to use some
direct measure of instruction time since it is possible for a machine to
have a fast access time and a much slower instruction time than comparable
machines. Add time is used since the type of circuit logic which makes
add slower or faster also makes most other instructions slower or faster.
In addition, two access-time instructions are very frequently used, and
add time by itself is not an unreasonable representation of computational
speed.
The term is used to allow consideration of the number of bits
p
transferred in in parallel (B) in the denominator and thus avoid the diffi-
culties involved in multiplying logarithms. T is in the denominator since
a smaller time Is better and this increases the size of the answer. As
T is divided by B, the result grows even smaller as B increases.
T is multiplied by A to remove any undue advantage which might accrue
p
to very cheaply built machines having a very fast transfer rate and some-
thing slow like a ripple-shift add logic. In addition, any slight advantages
in computational speed by one machine over another should be fairly portrayed,
since it is computation and not transfer rate that gets the task accomplished.
Table 3- 3 shews the machines evaluated by the Highland Method.
3-133
In the Highland method there are a number of improvements over the
other methods. An examination of Table 3-3 may rock some of our pre-
conceptions^ but reference to columns M, T, A and B will show why machines
are ranked as they are.
As with the Babbage, the resulting Highland number may be operated
upon arithmetically for purposes of solving analytical problems. This
may be done since the rating number scale, after having been both mul-
tiplied and divided, is now linear (or very close to it) instead of
logari thmic
The Highland Method measures what we wish to consider in a logical
and mathematically consistent manner. The resultant ratings may be
manipulated analytically. Finally, the analyst has a method for adjusting
the marginal value of incremental memory to the potential user for the
task at hand.
3.4.2.9 Conclusion
It must be understood that Figures of Merit have severe limitations
both in their field of application and in the scope of factors which they
consider. However, they are of great value to the analyst who understands
them thoroughly. They can be at the same time, professionally threatening
to the executive or administrator who uses them casually - - without an
understanding of what they mean or measure.
There is no satisfactory way at this time to bridge the gap between
having a data processing requirement and selecting the appropriate machine
for it, except to perform a detailed analysis of the task at hand. This
analysis will necessarily include a bench-mark analysis unless the require-
ments are well-known in relation to the capability of a particular computer.
Only then will a Figure of Merit comparison yield any meaningful results
directly. Even so, the next step is often a benchmark analysis.
The next limitation of Figures of Merit is that they necessarily
cannot evaluate input/output capability or peripheral equipment configuration
since these are system (or problem) oriented and cannot be adequately
determined in advance of problem definition.
Total Bits Mem.
(M)
Access Time
(T)
Microsecs.
Bi ts Transfe
in Parallel
3rred
(B)
Add Time
(A)
Microsecs.
log^oM
K log^QM
(K = 12)
(The
K log^QM
-I
i Highland Rating)
CDC 6600
15,720,000
.35
60
.7
7. 19645
86.3574
.00408
21,166
Philco 212
3,120,000
.75
48
.6
6.49415
77.8298
.00936
8,315
IBM 7030
16,768,000
1.10
64
1.5
7.22453
86.6944
.0256
3,386
Hughes
H-330
6,288,000
.90
48
1.8
6.79851
81.5821
.0337
2,420
SDS 9300
768,000
.87
24
1.75
5.88536
70.6243
.0633
1,115
RCA 601
1,792,000
.75
56
5.7
6.25334
75.0401
.0752
997
P-B 440
672,000
2.50
24
1.0
5.82737
69.9284
. 104
762
Univac
1107
2,340,000
2.00
36
4.0
6.36922
76.4306
.222
344
Univac
490
960,000
3.00
30
4.8
5.98227
71.7872
.480
149
CDC G-20
1,024,000
3.00
32
15.0
6.01030
72. 1236
1.40
51.5
SDS 910
384,000
4.00
24
16.0
5.54158
66.4990
2.67
24.9
CDC 160-A
384,000
3.20
12
12.8
5.54158
66.4990
3.42
19.4
HIGHLAND METHOD FIGURE OF MERIT (With K for»^.8 Value)
TABLE 3-3
CO
I
4^
3-135
Some additional key factors which are not considered by Figure of
Merit methods are; instruction repertoire, anwunt and type of low speed
storage, mean time between failure, mean time to restore, and amount of
memory cycle overlap. These factors must all be carefully weighed in
any complete analysis.
Figures of Merit may be used quite well to evaluate the relative
power of various central computers and their high speed memories indepen-
dent of their application to a specific problem. Not only can they be
used to solve the analytical problems posed earlier and other problems
closely related, but also they can be used very effectively to evaluate,
from a cost-effectiveness point of view, proposed changes to data pro-
cessing systems.
When memory size is considered, parity bits and sign bits should be
excluded from the total since they store little or no information.
Some are required but others may be superfluous for the task at hand.
The number (M) to be used is the largest memory size that the particular
machine can be expanded to.
The illusory potential of logarithmic scales was more than completely
covered in a previous section. This quality must always be kept in mind
by the analyst. It begins to fade as linearity is restored by operating
on the log arithmetically. Unintentional changing of the base of the
logarithm wi 1 1 result, however, if care is not exercised with these
manipulations.
Access Time and Cycle Time must be used carefully in evaluating
destructive and non-destructive readout machines.
Another effect must be guarded against, in some machines memory
banks may be arranged so that access time may be reduced by referring to
these banks in rotation. This Is called "overlapping". Some machines
have this capability - others do not. The amount of overlapping allowable
varies among models and as a function of how many blocks of memory are
purchased. Since the number of memory blocks to be required cannot often
3-136
(if ever) be accurately determined at this stage of analysis, overlapping
should be considered by the analyst; but not in the figure of merit
computation.
One of the very low access times quoted by one manufacturer results
from maximum overlapping (which cannot be used unless all possible memory
banks are acquired), while a very low access time quoted by another manu-
facturer can still be reduced to about 2/5 of that quoted by the use of
his maximum overlapping capabilityo So much for the technical content
of descriptive literature. The competent analyst must be certain where
each of his numbers came from and why.
Add time is probably as good an indicator of internal computational
speed as can be found, and using it alone does not inject the tincture of
simulation mentioned earlier. In certain situations where the internal
speed of the machine is quite critical, the Information Channel Capacity
technique should be considered. Often, however, the technique used in
the Highland Method should be adequate.
The concepts concerning word size have been treated adequately in
previous sections, but it is important to remember that big words are not
tantamount to better machines in all instances.
Since cost cannot accurately be predicted early in the analysis, and
since costs are subject to change due to the pressures of competition,
costs must remain outside the computation. This is true even though they
must be considered in any worthwhile analysis.
When only a very small proportion of the high speed memory of a
particular machine is of a very much higher speed than the balance, such
as 128 registers of thin film vs. 32 K registers of core, then the thin
film speed may be neglected entirely for the Figure of Merit computation.
However, if we begin to postulate machines which have 5-10% or more of
main memory operating ultra-high speed, then this clearly must be con-
sidered in the computation. Just how to do this best is open to discussion
at the moment. In the Highland Method this factor would likely appear as
some sort of multiplier in the denominator.
3-137
References - Calculation of Figures of Merit - Section 3.4
1. Rector, R. W. Measuring the Capability of Computing Equipment.
Private Communication - unpublished.
2. Picket, R. S., Investigation in Search of a Measure of Data
Processing, Unpublished, April 1962.
3. Campbell, E. K. , The Determination of the Meaningful N- Tuples of
Instructions in a Computer Program, TM-865, 30 Nov., 1962, The
System Development Corp., Santa Monica, California
4. Anon, Dynamic Instruction Count of a Real Time Program, IBM
Federal Systems Division, Kingston, N. Y. , 21 Oct., 1960.
5. Anon, Mathematical Models for Information Systems Design and
Calculus of Operations, Magnavox Research Laboratories, MRL
Report #R-451, 27 Oct. 1961.
4-1
k. STUDY INTEGRATION TASK
4.1 SCOPE AND OBJECTIVES OF STUDY INTEGRATION TASK
4clol Merge Outputs of Requirements, Technology, and Methodology
The ANTACCS Study Integration Task merges the outputs from the
Requirements, Methodology, and Technology tasks; and develops and
demonstrates a set of approaches to ANTACCS system design. The in-
tegration task examines the outputs from the other tasks with the purpose
of system synthesis, and evaluating and comparing alternatives. Among
the major parameters to be considered are:
1 ) Miss ions
2) Command Levels
3) Timel iness
4) Operational Tasks
5) Information Processing Tasks
6) Data Flow
7) Standard Operating Procedures
The inputs to Integration may be classified as follows:
1) From Requirements:
a) All mission requirements for Naval Tactical Command
Control systems for the 1970 - 1980 time period.
b) A substantial documentation of the interrelationships of
these mis ion requirements.
c) Projections of tactical concepts and procedures for the
missions and time period stated.
d) Projections of the naval forces available to carry out
these missions, their formations, numbers, etc.
4-2
2) From Technology:
a) Current and Projected Hardware and Software Technology
as it applies to information processing systems.
b) The implications of this technology to the design and
operation of Naval Tactical Command Control Systems.
3) From Methodology:
a) The methods for properly planning and designing an
information processing system for command control.
This will include explanations of or references to
appropriate numerical or analytical tools.
b) The methods and techniques which may be used to plan and
control the implementation of information processing
systems - and possibly by extrapolation - command control
systems .
c) Examinations of a few external areas of interest (such as
simulation, modeling, and simulation languages) which,
though not an integral part of design or implementation,
are of interest to command control personnel.
The aim of integration is not to develop a preliminary design of ANTACCS
but to illustrate the procedures and analytic techniques which can be
applied by Naval system planners who will be synthesizing and evaluating
alternate approaches to ANTACCS.
4cl.2 Comparison of Implications of Alternate System Operating Concepts
for ANTACCS
On the basis of the ANTACCS system requirements which characterize
the various system operating concepts, the Integration task describes
hardware/software implications and puts together a system configuration
for each operating concept. These alternate configurations will be
compared by applying the techniques and procedures developed and
specified by the Methodology task. The following comparison parameters
will be used :
^-3
1) Data processing equipment requirements
2) Software requirements
3) Manpower requirements including calibre of personnel required
and training requirements
k) System Costs (using the Total Force Cost Concept)
5) Intership Communication Requirements
6) Vulnerability to Natural Interference
7) Vulnerability to Man-made interference
a) Unintentional
b) Intentional
i) Active
i i ) Pass ive
i i i ) Spoof i ng
4.1.3 Demonstration of Application of Techniques and Procedures to the
Synthesis and Evaluation of a System Node
Part of the Integration effort is to demonstrate the application
of Methodology (techniques and procedures) to the synthesis and
evaluation of a system node. In the process, the information needs
and level of detail required for anlaysis becomes apparent. Because
the level of detail required for a complete analysis of a system node
is greater than for a system comparison, and this level of detail cannot
be provided for every node, the node to be analyzed will be selected
jointly by the Requirements and Integration tasks on the basis of
availability of detailed data to adequately describe a node, completeness
of nodal description and appropriateness of the node to ANTACCS.
Procedures to estimate system performance will be selected from
the methodology outputs, and applicable constants will be determined
by the Integration Task. Boundary conditions describing the nodal
requirements, inclusive of concurrency of missions, will be established
and used subsequently to estimate hardware/software/procedures configura-
tions which satisfy the ANTACCS operation requirements. Each hardware/
software/procedures configuration will be used as a basis for estimating
k-k
the Intra-node system implications and communication requirements.
System boundary conditions will be isolated from the system require-
ments by analyzing system stress conditions; timeliness, traffic
volume, processing routines, etc., and will be used to synthesize
alternate configurations. The trade-off parameters for estimating
system performance will be selected from the point of view of critical
system performance, and measurabi 1 i ty . Parameters that cannot be
measured or observed are useless in system design or evaluation. The
system performance characteristics will then be estimated by considering
the hardware/software/procedures configuration, the system boundary
conditions, and estimated value or range of values on the system
trade-off parameters. Estimated system performance characteristics
will be tabulated for quick comparison, and used to compare one system
against another.
ko] .k Discussion of System Planning Items
The integration task will also examine the items covered in a
TDP and will isolate the information developed by the ANTACCS study
that specifically applies to the development of an ANTACCS/ACDS TDP.
Topics to be included in this discussion are:
1) Management of the program
2) System design and specification
3) System test and evaluation
4.1.5 Summary of Study integration Tasks
In summary, the Study Integration task will:
1) Merge the outputs from the Requirements, Technology, and
Methodology tasks.
2) Demonstrate the application of these outputs to the synthesis
and comparison of the attributes of various system configurations
resulting from the system operating concepts as hypothesized
and described by the Requirements task.
4-5
3) Demonstrate the application of the outputs from the Require-
ments, Technology, and Methodology tasks to the synthesis and
evaluation of an ANTACCS system node.
k) Isolate and discuss the items developed by the ANTACCS study
which are directly applicable to the preparation and speci-
fication of an ANTACCS/ACDS TDP.
k-6
4.2 COMPARISON OF IMPLICATIONS OF ALTERNATE SYSTEM OPERATING CONCEPTS
4 » 2 . 1 General
A candidate system is a complex of men, hardware, and software
that functions in concept according to prescribed procedures to accomplish
given objectives. The prescribed procedures and objectives are embodied
in the descriptions of the input data, output data, data base, data
processing functions, decision points, and data displays. The require-
ments of a candidate system are completely defined by specifying the
parameter details within the above categories. However, the hardware-
software-human relationships (HSHR) are not determined. Indeed, only
the boundary conditions are delineated. The HSHR are functions of the
state-of-the-art in hardware and software, space and power limitations,
personnel limitations, economics, etc. It should be noted also, that
the candidate systems as defined apply equally well to a data display,
a ship, a level of command, or even a complete task force. Different
system operating concepts are reflected in the detail definition of
system parameters.
In this manner the Requirement task will define the command
structure for alternate system operating concepts by specifying the
levels of command, command activities and the interconnecting information
flow lines. Direction of flow, type of traffic, volume of traffic,
and timeliness of each type of traffic would be associated with each
information flow line.
The command structure will serve as a framework for developing
the mission descriptions and candidate systems. Each mission, its
objectives, functions, etc., will be defined by the Requirements
task for each command level, from the top command level which interfaces
with the strategic command net to the lowest level of command which
interfaces with the sensor and weapons systems.
h-1
^'2.2 ANTACCS Structure and Organization
4.2.2cl Objectives of Structuring and Organizing
The first step in the comparison of alternate system operating
concepts is the structuring of the ANTACCS information which characterizes
these operating concepts. The areas to be covered in the structuring
process are:
1) Structuring of the operational elements making up ANTACCS with
respect to command, mission, platforms, and operational
conceptSc
2) Identifying concepts of system operation"
3) Identifying the information sets needed by Integration to
perform the design function which brings together technology
and requirements.
k) Presenting a structure and procedure for developing an information
model for command and control which will assist in formulating
alternate system configurations, and will facilitate performing
trade-off analyses.
This process will show the direction to be taken and the information
requirements that will be needed to fulfill the basic project objective
of identifying and evaluating alternate approaches to system design of
ANTACCS. These alternate designs will be approached from the point of
V i ew of:
1) Postulating an operating concept (e.g., centralized,
decentral i zed) .
2) Representing the requirements of ANTACCS command levels.
3) Hypothesizing hardware/ software/man configurations for
elements or nodes of the ANTACCS system.
4-8
k) Evaluating each alternate configuration in terms of criteria
such as cost, capacity, capability, flexibility, etc.
Specifically, the nodes that would be included are those of the
four command levels of ANTACCS: Task Force Commander, Task Group
Commander, Task Unit Commander, and Task Element Commander.
4c 2. 2. 2 Platforms - Missions - Command
At any point in time, the U<.S. Navy has an aggregate of platforms
to perform the many missions within a current Naval command organization.
The platforms represent at least a twenty-year span of marine engineering
construction techniques and weapon technology. The missions include
the traditional Navy missions (of which AWS, AAW, STRIKE, and AMPHIB
will be the objects of ANTACCS study) along with the requirements of
responding to the threat of new technological weapon developments.
The naval command structure remains essentially the same, but the
information needs and response time requirements for different levels
of command vary widely with mission, weapons, and threat.
Responding to any and every threat is a formidable task that
requires experienced personnel, exercising timely and effective control
over the U. S. Navy units participating In assigned operations.
4.2.2.3 Platforms
The ANTACCS structure will be consistent with and conform to the
limitations of the existing and anticipated U. S. Naval platforms. The
primary classes of platforms to be considered in the development of
alternate configurations of command and control systems for the
ANTACCS study are:
Aircraft Carriers: CVA, CVAN and CVS
Command Ships: AGMR
Cruisers: CA, CAG, CG, CGN, CGL
Destroyers: DD, DDE, DDG, DDR, DL DLG, DLGN
Submarines: SS, SSG, SSGN
Aircraft: Airborne Command and Control Centers
4-9
Supporting platforms will be included in the alternate
system descriptions whenever necessary for completeness in the
analysis and evaluation. Classes of platform information
desired are: weapons, sensors, communications, mission performance
capabilities, command leveT associations, etc.
4.2.2.4 Missions
The ANTACCS study is directed to the analysis and evaluation
of an amphibious operation which includes a detailed analysis of the
following missions (and to others as they are determined to be
appropriate) .
Anti Air Warfare (AAW)
Anti Submarine (ASW)
STRIKE
AMPHIB
The information requ i rements f or each mission will be described
by BAARINC for every level of command within the purview of ANTACCS
to provide a basis for effectively integrating the existing and proposed
Naval subsystems (NTDS, ATDS, etc.) Into an integrated command and
control structure.
4.2.2.5 Command Structure
The U.S. Naval command structure has traditionally been super-
imposed on the platform oriented elements of the Navy which occurs
naturally because each ship can be considered as a single entity.
Because of the platform orientation, and the need for flexibility within
a naval organization, each ship must be capable of performing multiple
missions, both sequentially and concurrently. Sequentially, whenever a
4-10
primary mission is changed, and concurrently whenever multiple missions
are assigned. The multiple mission-multiple command level capability
of the ships provide great flexibility for making up a Task Force, for
mission assignment and re-assignment and for utilizing a Task Force for
new missions other than the ones for which it was initially formed.
The thread which coordinates and unites these diverse, mission-oriented
complexes of ships and men is the Naval command and control structure.
The basic command structure which will be used by the Integration
Task in synthesizing and evaluating candidate systems for ANTACCS is
illustrated in Figure 4.1 Naval Operational Chain of Command for
ANTACCS.
This structure illustrates six levels of command which extend from
the Fleet Commander in Chief down to the Element Commander. The lower
four levels of command represent the command span covered by ANTACCS.
ANTACCS will interface with the Numbered Fleet Commander above, and
will extend to the Element's Combat Information Center at the lower end
of the command structure.
ANTACCS will integrate the existing and proposed Naval weapon,
sensor, and command systems such as NTDS, MTDS, ATDS, SEAHAWK, CAPE,
FRISCO, lOIS, etc. into a unified Naval command structure within
which the future navy will operate. These systems will have a unique
interface within ANTACCS at each level in the command structure. Each
interface will be defined by data transfer and timeliness response
required for effective inter-system operation. Intuitively, one is
led to believe that each of these weapon and sensor systems will
interface within ANTACCS at several levels of command. For Instance,
the NTDS system will interface with ANTACCS on the element commander
level by providing raw and/or processed sensor data which reflects
dynamic environmental conditions, and by receiving weapon and sensor
assignments from the element commander.
4-11
FLEET
COMMANDER-
IN-CHIEF
I
NUIviBERED(TASK '
FLEET
GOMKANDEit
1
NUMBERED (TASK)
FLEET
COMMANDER
TASK FORCE
COMMANDER
TASK FORCE
COMMANDER
TASK GROUP
COMMANDER
TASK GROUP
COMMANDER
TASK UNIT
COMMANDER
TASK UNIT
COMMANDER
A
N
T
A
C
C
s
FUNCTIONS
Wpns.Cont. Nav,
Sens. Cont.
Other
Fig. 4-1: NAVAL OPERATIONAL CHAIN OF COMMAND FOR ANTACGS
4-12
NTDS will also Interface on a higher command level where decisions
affecting area of coverage and target responsibility between NTDS and
AIDS are to be made. The information necessary to support area of
coverage assignments certainly will not be raw sensor outputs, but
processed environmental summaries which reflect magnitude and direction
of raid instead of individual aircraft locationo Changes in area of
coverage will probably be based upon a number of factors, some of which
might be: size of raid, threat axis, probable enemy reserves, etc.
In conclusion, each level of command will have its unique interface
with the weapon and sensor systems, and will have its unique problems
associated with control and utilization of the weapon and sensor
systems .
The information requirements will be delineated by the breadth of
command responsibilities within each level of command, and by the
extent of direct control exercized over sensors and weapons » Breadth
of command is a function of the number and type of missions assigned,
and of the mission objectives. As a general rule, the greater the
breadth of command, the greater becomes the probability that only
summary data will be provided as a routine basis, with detailed
supporting data to be made available on short notice. Also, the larger
the number of simultaneously assigned missions, the more diverse and
less detailed becomes the information needed to support the commander
in the execution of the missions. In the development and evaluation
of the alternate system for ANTACCS, the information requirements for
each level of command and mission will be determined and compared with
the requirements for other missions. The flexibility of the ANTACCS
alternate systems and operational efficiency for multiple mission
performance will be estimated through the techniques of comparing
information, personnel, and equipment requirements for each level
of command.
4-13
4.2.3 Concepts of System Operation
4.2.3^1 General
Initially two philosophies of system control, central i zed and
decent ral i zed, wi 1 1 be analyzed and evaluated. While centralized and
decentralized control represent extremes in system control, a modified
and perhaps more realistic interpretation of each concept will be used
in the ANTACCS Analysis.
4.2.3«2 Centralized Control
The centralized system control concept is illustrated by Figure 4.2,
Centralized System Control. This figure is grossly simplified, however,
the concept of centralized control is clearly depicted. First, it is
assumed that each ship in the task force is capable of performing the
control function of each assigned mission, and that each ship can
satisfactorily perform the target detection and weapon control operations
associated with the assigned missionc In the figure. Ship number 3 has
been designated Force Anti Air Warfare Commander (FAAWC) for the task
forcec In this capacity, FAAWC would receive processed and semi processed
outputs from the AAW sensors throughout the task force, evaluate the
environmental data, and initiate management directives for the assign-
ments of AAW weapons to targets « Actual control of the weapons would
be exercised by Combat Direction Centers located on each ship. Since
each ship would be equipped to perform the AAW control function of the
Task Force, this function could be immediately transferred from one
ship to another, if necessaryo
A similar situation prevails for the Force Anti Submarine Commander
(FASWC) . The FASWC would receive processed and semiprocessed outputs
from the ASW sensors throughout the Task Force, evaluate the ASW
environment and initiate management directives for the assignment
of the ASW weapons to the targets. The FASWC would exercise manage-
ment control over the Task Force ASW sensors and weapons, but active
control would be exercised by the Combat Direction Centers on each shipo
4-14
AAW
Co n t ro 1
ASW
Control
TASK GROUP
AAW Info.
ASW Info.
^ o o o
Figure 4-2; CENTRALIZED SYSTEM CONTROL
i+-15
While each ship would be capable of assuming the Force mission
control function of each mission, this does not mean that the mission
control would be performed equally well by each shipo Differences in
available space, equipment, and personnel, would affect the efficiency
of mission control from one ship to another. Another factor which
would also influence the efficiency of mission performance would be the
number of concurrent missions which would be performed on a single
shipc Larger ships could and would probably perform more than one
concurrent mission. However, each ship would have a saturation point,
which is a function of the equipment, personnel, and data volume and
beyond which the system performance would begin to deteriorate rapidly.
The saturation point would probably differ for each ship because of the
normal variation in equipment and personnel, but the variations in
saturation point conditions should be small and should lie within a
yet undetermined but small range of parameter values.
4.2.3.3 Decentralized Control
The decentralized system control cpncept is illustrated by Figure k.3
Decentralized System Control. It is assumed that each ship would be
capable of performing the control functions for all missions simultaneously,
and one ship within each sector would be assigned total responsibility
for the section. Within the sector, Sector Warfare Commander (SWC)
would perform the management control function for all missions and
would initiate directives for assigning sensors and weapons to targetSc
Each SWC would inform other SWC of the current environmental conditions
on a timely basis through a routine interchange of information and
through special messages for handover of target assignment and weapon
control. The area of responsibility for a damaged or inoperable SWC
would be reallocated among the remaining SWC ' s to maintain total area
coverage. Enlarging the area of control might result in degradation of
system performance. Incremental increases in area of coverage and
number of targets would result in incremental degradation of the system
until a saturation point was reached. Beyond the system saturation
Sector 1
4-16
AAW, ASW,etc
Control
(Sec. 1)
AAw , ASVv, etc
Control
(Sec. 4)
Sector 4
/
/
/
Sector 2
SHIP i AA^,A3'^, etc
Control
(Sec. 2)
AA¥,ASW,etc '-
Control
(Sec. 3)
Sector "^
Communication
TASK GROUP
Fig. 4-3; DECENTRALIZED SYSTEM CONTROL
4-17
point, system performance degrades much more rapidly than the in-
cremental increase in system loadc The saturation point will probably
differ for each environmental condition and force configuration, but
the variations in the saturation point conditions should be small and
should lie within a yet undetermined but small range of parameter values.
4.2o3c4 Implications of Centralized/Decentralized Philosophy
There are several implications of the Centralized/Decentralized
control concepts which will be enumerated, but not evaluated at this
time.
1) Both concepts imply that a trade-off between computer
facilities and communication facilities will be effected.
That is, the input data to each level of command will be
summarized and edited before data will be transferred to
the next level of command, and the amount of data transferred
will be influenced by command level needs, communications
facilities, and data processing facilitieso
2) Control and Applications
a) Centralized control concept permits application of ships,
equipment, and personnel to missions for which they are
best suitedc
b) Decentralized control concept requires a j ack-of-al 1 -trades
ability for each command level.
3) Area of Responsibility
a) Area of responsibility is greater for each mission in the
centralized control concept which might introduce
problems with long range;, high volume communications
especially under ECM.
b) Area of responsibility is smaller for each mission in the
decentralized control concept which implies shorter range, high
volume traffic.
4-18
k) Centralized control concepts would probably require fewer
computers and displays per ship because fewer missions
would be simultaneously assigned to a single ship.
5) Decentralized control concepts would be more flexible
because each command level ship would be capable of
assuming and discharging the responsibilities of sector
control, which would include command and control of
mu 1 1 i p 1 e mi ss ions .
k.2.k Basis for Comparison of Alternate System Operating Concepts
k.2.k.] Attributes of Each System Operating Concept
The initial step in comparing the alternate system operating
concepts is to assemble and organize the information which charac-
terizes these conceptSo The assembly of this information will be by:
1) Inter-node information flow characteristics which will relate
mission, data quantity, direction of flow, timeliness criteria,
inter-ship distance, communication facilities, etCo
2) Command level and procedures which will relate mission,
operational tasks, data needs, decisions, timeliness
criteria, information processing tasks, etCo
3) Task force complement and disposition which will relate ship
type, command level, intership distances, communications
fac i 1 i t ies, etc <>
Another purpose of the analysis associated with this task, other
than organizing the data, is to isolate similarities and dissimilarities
among the various information characteristics of the various system
operating concepts. The similarities are representative of the invariant
attributes among the system operating concepts, and system design
solutions based upon these attributes should show a degree of consistency
The dissimilarities are representative of the variant attributes among
4-19
the system operating concepts, and system design solutions based upon
these attributes should be examined to determine their impact on the
systems. Variations in system performance and operation should be
primarily reflected in design solutions which satisfy the variant
system attributes.
This effort was not completed for this reporto It will be
fully discussed in the ANTACCS Study final Report.
4.2c4.2 Estimate Hardware/Software Implications
The variant and invariant attributes of the system operating
concepts will be translated into their hardware/software implications.
The analysis steps included in this subtask are briefly discussed
below, and will be fully discussed in the final report.
1) Estimate quantity of data to be stored on each command level
which would include: classes of data, quantity of data,
data base update cycle, data retrieval cycle, mission, and
command level .
2) Estimate information processing functions to be performed on
the various classes of data which would include; processing
functions, classes of data, quantity of data, processing
time, concurrence of processing, mission, and command level o
3) Develop general system hardware/software configurations for
each system operating concept.
4.2c4o3 Comparison of Alternate System Operating Concepts
The analysis steps to be included in this sub-task are briefly
discussed below and will be fully discussed in the final report.
1) Define the basis for comparison of alternate system
operating concepts.
a) Hardware/ software/procedures/ personnel
b) Cost
i+-20
c) Intership communications
d) Vulnerability to natural and man-made interference
2) Compare the alternate system operating concepts
This comparison will be made against the hardware/software
configuration developed in task 4.2.^o2(4) and comparison
parameters defined above.
3) Document the results of comparison of alternate system
operating concepts
These results will be discussed in the ANTACCS Study final
report .
4-21
4.3 '''"nONbTRATION OF THE SYNTHEolS AND EVALUATION OF A SYSTEM NODE
4c3.1 Nodal Development
A node is an element of ANTACC3 which can be described relative to
its position in the command structure, the communication system, the
platform which carries it and a set of special command and control
functions. A node is completely defined by characterizing its inter-
and intrj-nodal parameters. Inter nodal parameters include: infor-
mation movement between nodes, relationships among nodes, nodal inputs,
and nodal outputs. Intra-nodal parameters include: mission, platform,
operating concept, command level, technical functions, command elements,
and operational tasks. The structure for organizing these parameters
will be .1 multidimensional array as described in par igriph 4.3.2.
Multidimensional arrays will be used to organize the background infor"
mation into d suitable structure for analysis of Jternate system con-
figurations. Examples of these arrays to organize background information
are: platforms vs command levels, pl-.^tforms vs primary missions, plat-
forms vs weapons and sensors, command levels vs information requirements,
command levels vs operational tasks, etc. In addition to the background
information, the multidimensional arrays will also be used to organize
the mission requirements information for each level of command. The
structured information will then be used as a basis for synthesis and
evaluation of the alternate configurations for ANTACCS.
The methods of information structuring will be amenable to analysis,
within limits, by applying set theory. Such concepts as null sets,
congruence, and intersection can be usefully employed to isolate related
and non-related technical operations and data baseso The applications and
limitations of set theory to information structuring will be investigated
and used wherever set theory techniques are found to be applicable.
4.3.2 Information Structuring
Past experience in planning and implementing on-line, real time infor-
mation processing systems has led to various information structuring tools
to guide the transition from broad requirements to specific implementations
4-22
of hardware, software and user procedures. Continuing work in this
field will undoubtedly produce more techniques in the future, however,
no universal techniques or methodology exist in this problem area at
this time. These considerations, however, do not mean that the analyst/
planner should not attempt to use the most objective techniques available.
In recognition of the current state of information structuring methodology
and technology. Informatics has selected a structuring technique as a
starting point which has been used successfully in a recent on-line
information processing system analysis and design. The basis of this
technique is a structure which depicts the information movement and pro-
cessing in the system under analysis in a framework consistent with the
particular problem, yet independent of particular configurations, echelons
or geographic divisions of the system in the real world.
In ANTACCS this means that the structure should depict the current
(seconds, minutes) and historical (hours and longer) environmental infor-
mation movement and processing, and the command information movement and
processing within the system in a framework which is consistent with,
yet independent of, four variables: particular command level, mission,
platform, or operational concept.
A structure thus developed is applicable as an analytical tool for
integration of intra-nodal requirements under differing combinations of
the four variables and for isolating the necessary data processing and
display operations after successful integration.
A structure which meets this criterion is a multidimensional frame-
work which details system operational tasks on one axis, elements of
command and control on the second axis and a series of definable and,
to the extent possible, logically separable information processing
functions on the third axis. Each of the information processing functions
is defined by its inputs, outputs, historical data banks or criteria, and
internal processeso The functions thus defined are linked by information
flow from function to function, feedback among functions and recycling
of some or all functions. This linking allows the structure to depict
4-23
the dynamics of the decisions and information flow for each operational
task. Figure k-k is an example of the structure. One axis represents
some elements of command and control: estimate of the situation, develop-
ment of plan preparation and issuance of directives, and monitoring
planned actions. Another axis shows some examples of system operational
tasks; surveillance, weapons assignment, weapon employment and movement
of forces. The third axis shows the information processing functions as:
1) Data Collection and Classification:
a) Monitoring static and dynamic parameters which define the
enemies, political objectives, and military capabilities
and actionsc
b) Monitoring static and dynamic parameters which define our
own political objectives, military capabilities, and actions.
c) Monitoring static and dynamic parameters which define the
neutrals' political objectives and military capab i 1 i t i es o
d) Monitoring static and dynamic parameters which define the
weather, and other physical phenomenao
e) Segmenting the data collected into various definable classes.
2) Data Conversion and Selection:
a) Converting like data to common base (all time to Z-time,
locations to common, etCo)o
b) Separating pertinent mission data from the total input data.
3) Information Correlation and Significance Determination which is:
a) Combining information from within a source or from various
sources to structure a partial or complete description of
events or situationsc
b) Determining possible event or situation outcomes by comparing
current partial description of events and situations with
past events and situations (pred i ct ion) o
c) Determine the relative worth, redundance, conflict, and/or
importance of various information..
d) Feedback to data collection and selection to make a change
in emphasis or input needs.
4-24
System Operational tasks
Elements
of
Command
and
Control
Information
Frocesslnp"
Functions
Surveil-
lance
Weapon
Assign-
ment
Weapon
Emplo3^-
ment
Ko vement
Estimate of
Situation
Development
of Plan
! Preparation
i & Issuance
of
Directives
Monitoring
Planned
Actions
Event/situation
Relevancy
Determlna tlon
Information
Correlation
Significance
Determination
Data Conversion
and Selection
Data Collection
and Classification
/-—
/
Fip. 4-4; INTRA NODE DEFINITION; GIVEN KISSION, PLATFORM,
COMMAND LEVEL, OPERATIONAL CONCEPT
4-25
4) Event/Situation Relevancy Determi nation which is:
a) Relating events/situations within the environment to particular
task, etc.
The Complete Description for Each Function Should Include:
1) Technical Function Title including expected level of analysis
2) Input to Function
a) Qual i tati ve
i) types of input
(class definition down to specific formats, etc.)
i i ) source of i nput
(forward from previous function, feedback from other
functions, etc.)
b) Quantitative
i) frequency or occurence rate of inputs
ii) timeliness of input
(related to origination of data)
iii) concurrence of multiple sources
3) Historical and/or Criteria Files of Function
a) Qualitative (data type description)
b) Quantitative
i) amount of data
ii) update frequencies
c) Retrieval or use times/frequencies
4) Logical Processes within Function
a) Qual i tat ive
i) information/data relationships
(derivable or inferable classes)
4-26
i) logical algorithms
ii) use of historical data, criteria, or thresholds
v) decis ion resul ts
b) (Quantitative
) frequencies
i) concurrences
i i ) time delays
5) Outputs
a) Q.ual i tative
i ) types
i i ) formats
i i i ) routing
b) Quantitative
) lengths of messages, information groups, etc.
i) frequencies by types, etc.
ii) timeliness required and/or timeliness achievable by types
4-27
4.3o3 Structure Use
Structures developed In the pattern described above can be used to
illustrate the basic qualitative and quantitative aspects of information
flow and processing at many levels of detail within a node and many
levels of abstractness, depending on the stage of problem requirements
definition^ For instance, functions should be defined for a complete
task force engaged in every mission at a general level to establish a
point of reference for starting a detailed analysis, as well as
completely detailing the flow and processing that should occur for a
node at a particular command level (Task Group) performing a particular
mission (decentralized AAW) using particular platforms (DLG's and DDR's).
Once an intra-node structure has been defined, the data processing
operations can be definedc Some typical operations are:
1) input - on line from machine
- manual entries
2) output - on line displays
- printouts
- alarms
- to another machine on line
3) computing - arithmetic
- log ic
4) filing and collecting
5) sorting
6) file purg I ng
7) file search ing
This statement of the data processing operations which must be
performed, along with all the various file sizes and an indication of
the complexity of the logic, allows an estimate of the equipment
characteristics (dp and display and comm.)> program specifications, and
procedures necessary to meet the node requirements. Alternate projected
hardware, software, and procedure descriptions can now be assembled and
evaluated .
4-28
4.3.4 Example of Integra tion Approach
A preliminary illustration of the structuring technique described
in section 4.3.2 has been constructed using the BARRING TFC nodal
description. It is presented below under the two primary headings of
Intra-nodal parameters and inter-nodal parameters.
4.3.4.1 Intra-Nodal Parameters
1) Node to be Described
Command Level - TFC
Platform - Not Given
Mission - AAW (limited war)
Concept of Operation - Centralized Control
2) Elements of Command and Control
Estimate of the Situation
Development of the Plan
Preparation and Issuance of Directives
Supervision of Planned Action
3) System Operational Tasks
Allocation of Support Units
Relative Disposition of Forces (Position, Time, Movement)
Determination of Threat Axis, Magnitude
Surveillance of Current Force and Threat Status
4) Information Processing Functions
a) Estimate of the Situation, Relative Disposition of Forces,
Threat
i) Data Collection and Classification
Inputs - Intelligence Report, Movement Report, Own Forces
Outputs - Not Given
Processes - Not Given
Historical - Not Given
Banks or
Cri ter ia
4-29
ii) Data Conversion and Selection
Inputs - Not Given
Outputs - Not Given
Processes - Not Given
Historical - Not Given
Banl<s,Cr i ter ia
iii) Information Correlation and Significance Determination
Inputs - Not Given
iv) Outputs -
Printout - Enemy Forces Information
Printout - Friendly Forces Information
Plot - Enemy Forces Time and Position Plot
Plot - Friendly Supporting Forces Time and Position Plot
Printout - Relative Combat Power
Printout - Opposing Characteristics
v) Processes - Sort, Search
vi) Historical Banks, Criteria
Enemy Forces in Area of Operation
Friendly Supporting Forces in Area of Operations
Area of Operations Data
Area Communications Facilities
vii) Event/Situation Relevancy Determination
Inputs - Not Given
Outputs - Not Given
Processes - Not Given
Historical - Not Given
Banks
b) Development of the Plan, Allocation of Supporting Units,
Determination of Threat, Relative Disposition of Forces
i) Data Collection and Classification
Inputs - Component Operations, action statements,
threat axis, schedule, coordination instruction
4-30
Outputs " Not Given
Processes - Not Given
Historical - Not Given
Banks
li) Data Conversion and Selection
Inputs - Not Given
Outputs - Not Given
Processes - Not Given
Historical - Not Given
Banks
iii) Information Correlation and Significance Determination
Inputs - Not Given
Outputs - Not Given
Processes - Not Given
Historical - Not Given
Banks
iv) Event/Situation Relevancy
Inputs - Not Given
Outputs - Events by operation, forces, formation,
tactical net plan
v) Processes - Search
vi) Historical Banks "
Own Forces and Status
Own Forces Characteristics
Enemy Forces
Formations
Current Disposition
Future Events
c) Preparation and Issuance of Directives
Not Given
4-31
d) Supervision of the Planned Action, Surveillance
I) Data Collection and Classification
Inputs - Own Forces Status Position
Enemy Forces Status Position
Outputs - Not Given
Processes - Not Given
Historical - Not Given
Banks
li) Data Conversion and Selection
Inputs - Not Given
Outputs - Not Given
Processes - Not Given
Historical - Not Given
Banks
ill) Information Correlation and Significance Determination
Inputs - Not Given
Outputs - Not Given
Processes - Not Given
Historical - Not Given
Banks
Iv) Event/Situation Relevancy Determination
Inputs - Threshold Ps
Outputs - Time of Threat Onset
Significant forces change
Time Required to Change Disposition
Maximum Weapon Release Range and Time
Ps at Maximum weapon release and threshold test
Alerts
Processes - Not Given
Historical - Own Forces
4-32
Bvinks Current disposition
F.nemy Forces
Relative Combat Power
Format ions
4.3.4.2 Inter Nodal Parameters
1) Inputs - Not Given
2) Outputs - Not Given
3) Relationships Among Nodes - Not Given
NOTE: A node has not been selected for complete analysis to
date, however, the selection will be made soon and an analysis
presented in the final report.
4.305 Boundary Conditions
1) Statement of Bounds for each Node in the System
2) Technology Limits
3) Requirements Limits
NOTE: To be accomplished in final report to a detailed level
for one node as indicated and to a general level for all nodes.
4.306 Trade-Off Parameters
4.3.6.1 Operation Parameters
1) Level of Capability Achieved
2) Comprehensiveness of Node Solution
3) System Operation under Degradation Conditions
4) Time, Volume Capacities
5) Saturation Points
403.6.2 Implementation Parameters
1) Time to Implement and obtain operational capability
2) Complexity of implementation
3) Cost of subsystem and total
4) Cost of operating system (maintenance and supplies peculiar to
al ternative)
4-33
NOTE: The Trade Off Parameter definition will be improved and
figures provided in the final report for the node selected for
intensive study.
4.3.7 Descripti on of Hard ware/Software/Procedures Alternatives
1) Configuration Descriptions
2) Measures of Performance for Alternatives
3) Configuration Capabilities
Methods for analytically evaluating alternates of the complexity of
ANTACCS are yet to be fully developed, but will fall in the class of
techniques which assign relative values to the various criteria levels
achieved by alternates and form resultants for each candidate. The
resultants will then be compared using an appropriate numerical scale to
arrive at a best choice.
4-34
4.4 DISCUSSION OF SYSTEM PLANNING ITEMS
The present work by Informatics is intended to demonstrate, by
example for one node, the application of methodology and analytical
techniques to system design as a guide for the system planners who
will be responsible for complete system design and preparation of the
Technical Development Plan (TDP) for ANTACCSc With this view in mind.
Informatics has provided a discussion of several system planning items
which form part of a TDPo Items of particular interest are:
1) Narrative of Requirements and Brief of the Development Plan
2) Management Plan
3) Block Diagram of System
4) Subsystem Characteristics
5) Associated System Characteristics
6) Test and Evaluation Plan
Most of the work in the study integration task must necessarily
be done in the final months of the study effort. Accordingly the
discussion of this task in this midway report is of an introductory
nature only.
5-1
BIBLIOGRAPHY
5.1 INTRODUCTION
Bibliography is organized generaily in the sequence of the technical section
of the report. This section thus collects the bibliographical references
under the following main headings.
1) Technology
a ) Displays
D; input/Output Devices
c) Memories
d) Components and Packaging
e) Advance Usage Technique
f) Machine System Organization
g) Programming
2) Methodology
a) Computers and Hard Science
b) Simulation
At this point in the ANTACCS study, the bibliographical data has not
been completely cross-checked and merged. This will be done for the final
report.
5-2
5.2 TECHNOLOGY
5.2.1 Bibliography: Displays
Anderson, R. , A Synopsis of the State of the Art of Dynamic Plotting
Proiection Displays , Second National Symposium of the Society for In-
formation Display, New York, October, 1963.
Baron, P.C, Colordata: A Computer Driven Large Screen Display , Paper
presented to Orange County Chapter of lEEE-PTGEC, December 5, 1963.
Bauer, W. F. , and Frank, W. L. , DODDAC - An Integrated System for Data
Processing Interrogation and Display , Proceedings, Eastern Joint Com-
puter Conference, Washington, D. C. , December, 1961.
Blank, H. G. , O'Connell, J.A. , Wasserman, M. S. , Non-Linear Resistors
Enhance Display Panel Contrast , Electronics, August 3, 1963.
Bjelland, M. L. Epic D isplay , Proceedings, Third National Symposium on
Information Display, San Diego, Calif., Feb. 1964, pp. 286-299.
Darne, F.R. , Cathode-Ray Tubes , Electronic Information Display Systems,
Spartan Books, Washington, D.C. , 1963, pp. 87-109.
Davidson, R. A., and Helbig, W. A. Color Data Display , National Winter
Convention on Military Electronics, Feb 5-7, 1964, Los Angeles, Vol Ml,
pp. 14-2-14-14.
Haley, E. J., Photochromic Dynamic Display , Electronic Information Dis-
play Systems, Spartan Books, 1963, Washington D.C, pp. 110-120.
Harris, Lee T. , Status and Trends of Data Display Technology in Cornmand
and Control Systems , National Winter Convention on Military Electronics,
Feb 5-7, 1954, Los Angeles, Vol. Ill, pp. 14-1.
Howard, J. H. (Ed), Electronic Information Display Systems , Spartan
Books Inc. , 1963.
KIme, F. W. , and Hartley-Smith, A., Data Display System Works in Micro -
seconds, Electronics , McGraw Hill publication, November 29, 1963, Vol. 36,
No. 48, pp. 26-30.
Kulcke, W. , and Harris, T. J., Kosanke, E. , Max, E. , A Fast, Digital -
Indexed Light Deflector , IBM Journal, January, 1964, Vol. 8, No. 1,
pp. 64-67.
Lindberg, Evert, Solid Crystal Modulates Light Beams , Electronics,
McGraw Hill Publication, Dec. 20, 1963, pp. 58-61.
5-3
Loewe, R. T. , ARTOC Displays , Electronic Information Display Systems, Spar-
tan Books Inc., 1963, Washington, D. C. , pp. 231-246.
Loewe, R. T. , Sisson, R. L. and Horowitz, P., Computer Generated Dis-
plays , Proceedings of the IRE, January, 1961, Vol. 49, No. 1, pp. 185-195.
Lovell, Ron, New Displays for Space Flight , Electronics, McGraw Hill
Publication, Feb. 21, 1964, pp. 42-43.
Merel , W. , and Barkan, H. , Computer Compatible Electroluminescent Tech -
niques for the Achievement of Wide Angle Visual Displays , 1963 IEEE
International Convention Record, New York City, Part 4, March 28, 1963,
pp. 11-18.
Redman, J. H. , Advanced Display Techniques Through the Chractron Shaped
Beam Tube , Society for Information Display Symposium, March, 1963.
Rome Air Development Center, Criteria for Group Display Chains for the
1962-65 Time Period , Technical Documentary Report No. RADC-TDR-62-31 5,
July, 1962, pp. 102.
Second National Symposium on Information Display, Proceedings, October
3-4, 1963.
Smith, Sidney L. , Visual Displays- Large and Small . Mitre Corp., Nov-
ember, 1962, (ASTIA No. AD 293-826).
Society for Information Display, Proceedings, First National Symposium
on Information Display, March 14, 1963, Los Angeles, California.
Talmadge, H. G. Jr. , Physical Principles of Displays - Classification ,
Electronic Information Display Systems, Spartan Books, Washington, D.C.
1963, p. 69-86.
Thompson Ramo-Wol Idr idge Corporation, DODDAC, Advanced Operational Sys-
tem - Final Design Report , Contract DA-49-146-XZ-103, Report C153-2S-30,
Vol. 1 and II, classified SECRET.
Wasserman, M. S. , Display Appl ications of Electro-luminescence . Electro-
nic Information Display Systems, Spartan Books, 1963, Washington, D.C,
pp. 121-128.
White, G. R. , Review of Laser Applications , 16th Annual NAECON, May 12,
1964, Dayton, Ohio.
Yardo, Stephen, Sol id State Display Device . Proceedings of the IRE,
December, 1962.
5-4
5.2.2 Bibi ioqraphy; Input/Output Devices
American Standards Association, Minutes of the ASA Committee X3 . 1
(Optical Character Recognition) and its subcommittees, Sectional
Committee X3 on Computers and Data Processing.
Athens, A. S., Using Solar Cells to Read Holes , Electronic Design,
February 1962, 10:78-81 .
Barbeau, R. A. and Aweida, J. I.., IBM 73^0 Hypertape Drive . Proceedings
Fall Joint Computer Conference, Las Vegas, Nevada, November 12-14, 1963,
Vol c 24, pp. 591-602.
Beall, W. R., Tape Printer Applications , Instruments and Control Systems,
Vol . 32, pp. 708-709.
Blasbalg, H. and Van Blerkom, R., Message Compression , IRE-Transact ions
on Space Electronics and Telemetry, September 1962, 8:228-38.
Burr, R. T., The Printed Motor - A New Approach to Intermittent and
Continuous Motion Devices in Data Processing Equipment , Proceedings
Eastern Joint Computer Conference, New York, N. Y., December 13-15, I960,
Vol . 24, pp. 325-342.
Carroll, J. M., System Reads Three Type Fonts , Electronics, McGraw Hill
Publication, December 20, 1963, pp. 49.
Chapman, D. W., Optimizing the Digital Magnetic Recording Process (Letter),
IEEE--Proceedings, January 1963, 51:247-48.
Clapper, G. L., Digital Circuit Techniques for Speech Analysis . IEEE--
Transactions on Communications and Electronics, May 1963, 66:296-304.
Crowson, H. L., Error Analysis in the Digital Computation of the
Autocorrelation Function (Letter), AIAA Journal 1 (2) , February 1 963 ,
488-89.
Datamation, Electronic Retina Character Reader , July 1963, Vol. 9,
No. 7, PPo 50.
Datamation, National Cash Register Magnetic Matrix Printer , July-August 1958
Desblache, A., Data Treatment Using Numerical Transmission Over Long
Distances , Onde Electrique, February 1963, 431:243-50.
Dinneen, G. P., Programming Pattern Recognition , Proceedings Western
Joint Computer Conference, Los Angeles, California, March 1-3, 1955«
5-5
DIttmann, G. W., Introduction to Navy Tactical Data Systems , National
Convention on Military Electronics, Washington, D.C., September 11, 1963»
Epstein, Herman, The Electroqraph ic Recording Technique , Proceedings
Western Joint Computer Conference, Los Angeles, California, May 1-3, 1955»
Fan, G., Donath, E., Barrekette, E. S= and Wirgin, A., Analysis of a
Magneto-Optic Readout System , IEEE, Transactions on Electronic
Computers, February 1963, 12:3-9.
Freeman, Mo E., and Gilmore, J. C, Open Loop Digital Hydraulics Position
Computer Memory Arm , Hydraulics and Pneumatics, November 1962, 15:92-950
Friedman, C. V., On the Choice of Binary Codes and Thresholds . IEEE--
Proceedings, March 1963, 51(3):^78.
General Electric Review, Thermo-maqnetography , July 1952.
Hess, Herman, A Comparison of Discs and Tapes , Communications of the
ACM, October 1963, Vol. 6, No. 10, pp. 63^-638.
Hess, Herman, A Comparison of the Characteristics of Modern Discs and
Tapes , Discfile Symposium, Ho 1 1 y wood , California ( Informat ics) ,
March 6-7, 1963.
Holmes, W. S., Leland, H. R. and Richmon, G. E., Design of a Photo
Interpretation Automation , Proceedings, Fall Joint Computer Conference,
December 1962.
Innes, Frank E., High-Speed Printer and Plotter , Proceedings Eastern
Joint Computer Conference, New York, N. Y., December 13-15, I960,
Vol . 18, pp. 153-160.
Instruments and Control Systems, Digital Printers , Editorial Survey,
Vol. 32, May, 1959, pp. 700-707.
Jory, John H. , Hot Wire Anemometer Paper Tape Reader . Proceedings
Eastern Joint Computer Conference, New York, N.Y. , Dec. 13-15, 1960,
Vol. 18, pp. 276-278.
Kleist, R. A., et al. Single Capstan Tape Memory , Proceedings Fall
Joint Computer Conference, Las Vegas, Nev. , Nove. 12-14, 1963, Vol.
24, pp. 555-576.
5-6
Lindner, K. , Punched Card as Information Carrier and as Technical Prob -
lem , Feinwerktechnik, 1963, 67(2): 55-61.
Martin, V. C. , Printed Circuit Motors for High-Speed Incrementing of
Inertia] and Dissipative Loads , IEEE — Transactions on Industrial Elec-
tronics 10(1): 28-45, May, 1963.
McCormick and Paget, Printing Equipment for Medium. Intermediate, and
Large Size Computers , Staff of Cresap, Control Engineering, Jan., 1962,
pp. 91-95.
Peterson, W, W. , Error Correcting Code , M. I.T. Press, and John Wiley
& Sons , Inc. , 1961 .
Petrick, S. R. , and Willett, H. M. , A Method of Voice Communication
With a Digital Computer , Proceedings Joint Computer Conference,
Dec. 13-15, I960, New York, N.Y., Vol. 18, pp. 11-24.
Preisinger, M. , Xerography — A New Non-Mechanical Printing Method ,
Elektronik 12(2): 33-36, Feb. 1963.
Richards, R. K. , Digital Computer Components and Figures , D. Van Nos-
trand & Co. , Inc. , 1957.
Rusch, A. , High Stability Magnetic Tape for Data Processing Systems ,
El ectro- Technology , Dec, 1963, pp. 91-96.
Shew, L. F. , Discrete Tacks for Saturation Magnetic Recording, IEEE —
Transactions on Broadcast and Television Receivers, May, 1963, 9:56-62.
Shew, L. F. , High Density Magnetic Head Design for Noncontact Recor -
ding , IRE — Transactions on Electronic Computers, Dec, 1962, 11: 764-73.
Sims, John C. . Magnetic Reproducer £■ Printer , Proceedings Western Compu-
ter Conference, Los Angeles, California, Feb. 4-6, 1953.
Singer, R. J., A Self-Organizing Recognition System , Proceedings Western
Joint Computer Conference, Los Angeles, California, May 9-11, 1961, Vol.
19, pp. 545-554.
Stone, J. J. , Production of Magazine Labels by Videograph Process, Klyce ,
Proceedings Western Joint Computer Conference, May 3-5, 1960, Vol.
17, pp. 371-382.
Tyrrel 1 , D. H. , et al , Evolution of Digital Magnetic Tape Systems for
Use in Military Environments, Proceedings Fall Joint Computer Conference,
Las Vegas, Nev. , Nov. 12-14, 1963, Vol. 24, pp. 577-590.
5-7
Wa 1 ton , C . A . , Analog Input and Output System for a Real-Time Process
Control Computer System , Joint Automatic Control Conference — Proceedings,
June, 1962, 13, 4. 1-4,6,
Wier, J.M. , Digital Data Communication Techniques , Proceedings of the
IRE, Jan. 1961, Vol. 49, No. 1, pp. 196-209
5-8
5.2.3 Bjb] iogr^phy; hgmories
Allen, R. J., Superconductive Delay Line Memory , Proceedings, Military
Electronic Conference, Washington, D.C., Sept. 10-11, 1963, PP- 370-372.
Amemiya, H., et al, High-Speed Ferrite Memories , Proceedings Fall Joint
Computer Conference, Philadelphia, Pa., December 1962, Vol. 22, pp. 184-196.
Analex Corp., Analex Model 800 Random Access Disc File , Sales Brochure,
1963, No. U 1063-
Analex Corp., Analex Random Access Memory Systems , Sales Brochure,
1963, Noc U 1 163-
Angel, A. M., Symposium on Large Capacity Memory Techniques , The NCR
Magnetic Card Random-Access Memory, Macmillan, New York, 1962, pp. 149-162.
Baker, W. A., The Piggyback Twistor--An Electrically Alterable Nondestructive
Read-out Twistor Memory . Proceedings Intermag Conference, Washington, D.C.
April 1964, pp. 8-5-1 .
Barkouki, M. F. and Stein, I., Theoretical and Experimental Evaluation of
RZ and NRZ Recording Characteristics , IEEE Transactions on Electronic
Computers, April 1963, Vol EC-12, No. 2, pp. 92-100.
Barrett, W. A., A Card-Changeable Permanent-Maqnet-Twi stor Memory of Large
Capac i ty , IRE Transactions on Electronic Computers, 3 September 1961,
Vol . EC- 10, pp. 451-460.
Bartkus, E., Brownlow, J., Crapo, W., Elfant, R., Grebe, K., and Gutwin, 0.,
An Approach Towards Batch Fabricated Ferrite Memory Planes , IBM Journal of
Research and Development, April 1964, Vol. 8, No. 2, pp. I7-I76.
Bates, A. M. and D'Ambra, F. P., Thin Film Memory Drive and Sense Techniques
for Realizing a I67 Nsec Read/Write Cycle . Digest of Technical Papers,
Solid State Circuits Conference, Philadelphia, Pa., Feb. 19, 20, 21, 1964,
pp. 106-107.
Beck, E. R., et al, Tunnel Diode Storage Using Current Sensing , Proceedings
Western Joint Computer Conference, Los Angeles, Calif., May 9-11, I96I,
Vol . 19, pp. 427-442.
Bittman, E. E., The future of Thin Magnetic Films. Large Capacity Memory
Technigues for Computing Systems . Macmillan Publishing, New York, 1962,
pp. 41 1-420.
Bloom, L., Pardo, I., Kenting, W., and Mayne, E., Card Random Access
Memory (CRAM): Functions and Use . Proceedings Eastern Joint Computer
Conference, Washington, D.C, Dec. 12-14, 1961, pp. 147-157.
5-9
Bremer, J. W., Cryotron Computer Techniques , Pacific Computer Conference
IEEE, Pasadena, California, March 15-16, 1963, pp. 42-44.
Briggs, Go R. and Sarnoff, D., M i crocore-Backward Diode Shift Register .
1964 Intermag Conference Proceedings, Washington, D.C. April 1964, pp. 11-2-1,
Brown, J. N., and Newhall, E. E., The Storage and Gating of Information
Using Balanced Magnetic Circuits , 1964 Proceedings Intermag Conference,
Washington, D.C, April 1964, pp, 11-21.
Brown, Reese, Magnetic Films for Digital Computers . 1963 Pacific Computer
Conference IEEE, Pasadena, California, March 15-16, 1963, pp- 45-46.
Bryant Computer Products, Modular Mass Memory . Sales Brochure, 1962, B-627»
Bugler, R. Jo, Random Access File System Design . Datamation, December 1963,
Vol c 9, No. 12, pp. 31 .
Burne, D. L., et al. Large Capacity Memory Techniques for Computing
Systems , Coincident Current Superconductive Memory, Macmillan, New York,
1962, pp. 421-440.
Burns, L. L., et al, A Large Capacity Cryoelectric Memory With Cavity
Sensing , Fall Joint Computer Conference, Las Vegas, Nevada, November 12-14,
1963, Vol . 24, pp. 91-100.
Burns, L. L., Jr., Alphonse, G.A. and Leek, G. W., Coinc ident-Current
Superconductive Memory , IRE Transactions on Electronic Computers,
September 1961, Vol. EC-10, No. 3, pp. 438-446.
Burroughs Corp., On-Line Discfile System for Data Storage and Data
Commun icat ions . Sales Brockure 10001, 1963.
Campbell, S. G., Systems Implications of New Memory Developments . Proceedings
Fall Joint Computer Conference, Las Vegas, Nevada, November 12-14, 1963,
Vol o 24, pp. 473-480.
Carlson, C. 0., Grafton, D. A., Tauber, A. S., The Photochromic Micro-
image Memory . Symposium on Large Capacity Memory Techniques for Computing
Systems, May 1961 .
Carothers, J. D., et al, A New High Density Recording System : The IBM 1311
Disc Storage Drive with Interchangeable Disc Packs, Proceedings Fall
Joint Computer Conference, Las Vegas, Nevada, November 12-14, 1963, pp. 327-340
Carvicker, R. W., UN I VAC Fastrand Mass Storage - A UNIVAC 490 Subsystem,
Proceedings First Discfile Symposium, Los Angeles, Calif., March 1963
5-10
Carver, W. W., Comparing Storage Methods , Burroughs, Electronic Industries,
August 1962, Vol. 21, pp. 120-130.
Chamberlain, D. M., Transfluxors , RCA Technical Bulletin, October 1962.
Chang, H., A Synopsis of Magnetic Memories . 1964 Proceedings Intermag
Conference, Washington, D.C., April 1964, pp. 5-2-1.
Chang, C, and Fedde, G., Magnetic Films - Revolution in Computer Memories ,
Proceedings, Fall Joint Computer Conference, 1962, pp. 213-224.
Clapp, L. C, High Speed Optical Computers and Quantum Transition Memory
Devices , Proceedings Western Joint Computer Conference, Los Angeles,
California, May 9-11, 1961, Vol. 19, pp. 475-489.
Cohen, Mo L., Slade, A. E., and Varteresian, A Cryotron Multi -Level
Logic and Memory Circuit , Digest of Technical Papers, 1964 Solid State
Circuits Conference, Philadelphia, Pa., February 19, 20, 21, 1964,
pp. 102-103.
Cohen, Martin L., Cryotron ics-Probl ems and Promise , Proceedings Fall
Joint Computer Conference, Philadelphia, Pa., December 1962, Vol. 22,
pp. 232-233.
Coil, E. A., A Multi Addressable Random Access File System , I960 IRE
WESCON Convention Record, August I960, Part 4, pp. 42-47-
Coil, E. C., and Goodman, S. A., Librascope Mass Memory--A "Working"
Storage System , Preprints of papers presented at Informatics Discfile
Symposium, Hollywood, California, March 6-7, 1963.
Corneretto, A., Associative Memories , Electronic Design, Feb. 1, 1963,
Vol . II, No. 3, pp. 40-55.
Crowther, T. S., High Density Magnetic Film Memory TechnJ.gues , Proceedings
Intermag Conference, Washington, D.C., April 1964, pp. 5-7-1.
Dal ton, M. M., HCM-202 Thin Film Computer , Proceedings, Spaceborne
Computer Conference, Anaheim, California, October 30-31, 1962.
Danylchuk, I., Gianola, U. F., Perneski, A. J., and Sagal, M. W,
Plated Wire Magnetic Film Memories . 1964 Proceedings Intermag Conference,
Washington, D.C., April 1964, pp. 5-4-1.
5-11
Datamation, New Dlscfiles from Burroughs . June 1963, Vol. 9, No. 6, p. kS.
Datamation, IBM ]kkO , November 1962, Vol. 8, No, 11, p. 76.
Data Products Corp., Technical Data on the Discfile . Technical Brochure,
August 1962.
Davies, Paul M., A Superconductive Associative Memory , Proceedings Spring
Joint Computer Conference, San Francisco, May 1-3, 1962, Vol. 21, pp. 79-87
Davies, Po, The Associative Computer , Proceedings, 1963 Pacific Computer
Conference, Pasadena, California, March 15-16, 1963*
Davis, J. S. and Wells, P. E., Investigation of a Woven-Screen Memory
System , Proceedings, Fall Joint Computer Conference, Las Vegas, Nevada,
November 12-14, I963, Vol. 24, pp. 311-326.
Dodson, G. A. and Ruff, J. A., Charge Storage Diode for Magnetic Memory
Appi icat ions . Digest of Technical Papers, 1964 Solid State Circuits
Conference, Philadelphia, Pa., February 19, 20, 21, 1964, pp. 104-106.
Electronic News, New Discfile by Burroughs , April 22, 1963.
Electronic News, New IBM Storage Systems Holds One Billion Characters .
October 21 , 1 963 .
England, W. A., Miniature Computer Designed for Space Environments ,
Proceedings, Spaceborne Computer Engineering Conference, Anaheim, Calif.
October 30-31, 1962, pp. 95-101.
Falkin, Joel and Savastano, Jr., Sal, Sorting with Large Volume. Random
Access. Drum Storage . ACM Sort Symposium, Princeton, New Jersey,
November 29-30, I962.
Fortin, E. G., and Lessoff, H., Wide-Temperature Lithium Ferrite Cores
For Coincident-Current Memory Arrays . RCA Technical Bulletin, April 1964.
Fuller, R. H. and Estrin, G., Some Applications for Content-Addressabl e
Memor ies . Proceedings Fall Joint Computer Conference, Las Vegas, Nevada,
November 12-14, 1963, Vol. 24, pp. 495-508.
Futami, K., Oshima, S., Kamibyashi, T., The Plated-Woven Wire Memory
Matrix , 1964 Proceedings Intermag Conference, Washington, D.C. April 1964,
pp. 5-1-1.
5-12
Goetz, Martin A., Organization and Structure of Data on Discfile Memory
Systems for Efficient Sorting and Other Data Processing Programs . ACM
Sort Symposium, Princeton, New Jersey, November 29-30, 1962.
Go Ids tick, G. He and Klein, E. F., Design of Memory Sense Amplifiers.
IRE Transactions on Electronic Computers, April 1962, Vol <. EC-11, No, 2,
pp. 236-252.
Gratian, Jo W. and Freytag, R. W., Ultrasonic Approach to Data Storage .
£lecttonics, May 4, 1964, Vol. 37, No. 15, pp. 67-72.
Gross, W. A., A Gas Film Lubrication Study - Part 1 - Some Theoretical
Analyses of Slider Bearings . IBM Journal of Research & Development,
July 1959, Vol. 3, pp. 237-274.
Hagedorn, F. B., Some Principles and Properties of Superconducting Thin
Film Computing Devices . 1964 Proceedings Intermag Conference, Washington,
D.C., Apri 1 1964, pp. 1 .3.1 .
Halaby, S. A., Gregor, L. V., Rubens, The Materials of Thin Film Devices .
Electro-Technology, Conover-Mast Publication, September 1963, pp. 95-122d.
Haughton, K. E., Air Lubricated Slider Bearings for Magnetic Recording
Spacing Control. Large-Capacity Memory Technigues for Computing Systems .
The MacMillan Company, New York, 1962, pp. 3^1-350.
Hillegass, J. R. and Stratland, N., Random Access Storage Devices .
Datamation, December 1963, Vol. 9, No. 12, pp. 34-45.
Hoagland, A. S., A High Track-Density Servo-Access System for Magnetic
Recording Disc Storage . IBM Journal of Research g- Development, October I96I,
Vol . 5, pp. 287-296.
Hoagland, A. S. and Bacon, G. C, High Density Digital Magnetic Recording
Technigues . Proceedings of the IRE, January I96I, Vol. 49, No. 1, pp. 258-267.
Hoagland, A. S., Mass Storage . Proceedings, IRE, May 1962, Vol. 50, pp. 1087-1092
Hobbs, L. C, Comparison: Major Types of Mass Memories . Data Systems
Design, January 1964, Vol. 1, No. 1, pp. 16-21.
Hobbs, L. C, Review and Survey of Mass Memories . Proceedings, FJCC,
Las Vegas, Nevada, November 12-14,1963, Vol. 24, pp. 295-310.
5-13
Honeywell Sales Brochure, High-Speed Random Access for the Honeywell 400 .
1963, DP 2078(DSA 66A, 750663).
Hubbard, George U., Some Characteristics of Sorting in Computing Systems
Using Random Access Storage Devices , ACM Sort Symposium, Princeton, N.J.,
November 29-30, 1962.
IBM Reference Brochure 822 6595 1, IBM 1301 Disc Storage for the 7090
Data Processing System , 1961.
IBM Sales Brochure 520 1795, IBM 1302 New Horizons in Random Access Data
Processing for Manufacturing Industries , 1 963 •
IBM Systems Reference Library Brochure, IBM l440 Systems Component
Description 1311 Disc Storage Drive , 1962, No. A26 5668 0.
Informatics Inc., Preprints of Papers Presented at First Discfile
Sympos i um , March 1963°
Ittner, III, The Case for Cryotronics? , Proceedings Fall Joint Computer
Conference, Philadelphia, Pa., December 1962, Vol. 22, pp. 229-231.
Jack, R. W., Groom, R. G. and Gleim, R. A., Engineering Description of
the Burroughs Discfile , Proceedings, FJCC, November 1963, Vol . 24,
pp. 341-350.
Jacoby, Mo, A Critical Study of Mass Storage Devices and Technigues
with Emphasis on Design Criteria , IRE PG MIL, National Winter
Convention on Military Electronics, 1962.
Kaufman, B. A. and Hammond, J. S. Ill, A High-Speed Direct-Coupled
Magnetic Memory Sense Amplifier Employing Tunnel-Diode Discriminators ,
IEEE Transactions on Electronic Computers, June 1963, Vol. EC-12, No. 3,
pp. 282-299.
Kaufman, B. A. and Ulzurrun, E., A New Technique for Using Thin Magnetic
Films as a Phase Script Memory Element , Proceedings Fall Joint Computer
Conference, Las Vegas, Nevada, November 12-14, 1963, Vol. 24, pp. 67-76.
Kilburn, T., et a1. One Level Storage System , University of Manchester,
IRE Transactions on Electronic Computers, April 1962, Vol. E-C-11,
pp. 223-236.
5-14
King, Claude F., Factors Affecting Choice of Memory Elements . Proceedings
Western Joint Computer Conference, Los Angeles, California, May 9-11, 1961,
Vol . 19, pp. 405-410.
Koerner, R. J. and Searbrough, A. D <. , Theory, Organization, and Performance
of a Search Memory . Local Symposium on Search Memory, Los Angeles District
of IEEE, May 26, 1964.
Kompass, E. J., Mold Memories on Mesh. Control Engineering, McGraw Hill
Publication, March 1964, pp. 28.
Kriessman, C. J., Matcovich, T. J., Flannery, W. E., Low Power Thin Film
Memory . Intermag Conference 1963 Proceedings, 1963, pp. 3-3-1 - 3-3-7.
Kulcke, W., et al, A Fast. Dig i tal - 1 ndexed Light Deflector . IBM Journal,
January 1964, Vol. 8, No. 1 pp. 64-67.
Kump, H. J. and Speliotis, D. E., Fundamental Criterion for Recording on
Magnetic Surfaces . Proceedings 1964 Intermag Conference, Washington, D.C.,
April 1964, pp. 3-2-1 .
Kuttner, P., The Rope Memory: A Permanent Storage Device . Proceedings
Fall Joint Computer Conference, Las Vegas, Nevada, November 12-14, 1963,
Vol . 24, pp. 45-58.
Leaycraft, E. C. and Melan, E. H., Characteristics of a High-Speed
Multipath Core for a Coincident-Current Memory , IRE Transactions on
Electronic Computers, June 1962, Vol. EC-11, No. 3, pp. 405-409.
Leeber, R. R. and Lindquist, A. B., Associative Memory with Ordered
Retrieval , IBM Journal of Research, January 1962, 6.1, pp. 126-136.
Lee, E. S., Associative Techniques with Complementing Fl ip-Flops .
Proceedings Spring Joint Computer Conference, Detroit, Michigan, May 1-3,
1963, Vol. 23, pp. 381-394.
Lee, E. S., Solid State Associative Cells . 1963 Pacific Computer
Conference, IEEE, Pasadena, California, March 15-16, 1963, pp. 96-108.
Lennor, W. T., Jr. and Jordon, W. F., Auxiliary Memory Speeds Information
Retrieval . Computer Control Co., Electronics, May 11, 1962, Vol. 35,
pp. 102-104.
Lemaire, Dr. H., New Technigues for Ferrite Nanosecond Memories .
RCA Technical Bulletin.
5-15
LeVezu, C .. A Mul t iaperature Diqital Memory Having Nondestructive Sensing .
Proceedings Spaceborne Computer Engineering Conference, Anaheim, Calif.,
October 30-31, 1962, pp. 65-68.
Lewin, M. Ho, et al. Fixed Associative Memory Using Evaporated Organic
Diode Arrays , Pro Fall Joint Computer Conference, Las Vegas, Nevada,
November 12-14, 1963, Vol <> 2k, pp. IOI-IO60
Librascope Division of General Precision, Inc., LI5OO. Sales Brochure
G3-4139 . 1963.
Lohan, F. J., Criteria for Selecting Random Access Mass Memories , Data
Systems Design, January 1964, Vol o 1, No. 1, pp. 25-29.
Long, T. R., Journal of Applied Physics 30 . I960, pp = 1235.
Matcovich, T. J., Flannery, W., Adomines, A., Liciw, Wo, The Design of
an Evaporated Memory System , UNIVAC, Division of Sperry Rand Corp.,
ISSCC #69.
Maeda, H., Matsushita, A., Woven Thin-Film Wire Memory , 1964 Proceedings
Intermag Conference, Washington, D.C. April 1964, pp. 8-1-1.
McGee, W., Co, Effect of High Speed Memory Organization on Average
Instruction Execution Time , Fifth Annual Los Angeles Area Technical
Symposium, Hotel Statler, October 29, 1962.
McLaughlin, H. J., Discfile Memories, Instruments Control Systems ,
November I96I, Vol. 34, pp. 2063-2068.
McQuillan, J. D. R., The Design Problems of a Megabit Storage Matrix for
Use in a High-Speed Computer , IRE Transactions on Electronic Computers,
June 1962, Vol. EC-11, No. 3 pp. 390-4o4.
Meier, D. A., A Five-Megacycle Dro Thin-Film Rod Memory , National Cash
Register, Hawthorne, California.
Meier, D. A., Magnetic Film Rods Provide High Speed Memory , Electronics,
February 2, 1962.
Meier, D. A. and Kolk, A. J., The Magnetic Rod - A Cylindrical Thin Film
Element , Large Capacity Techniques for Computing Systems, Macmillan
Publishing, New York, 1962, pp. 195-212.
5-16
Morse, D. C, et al.. Logic Organization of the UNIVAC ADD-IOOO Aerospace
Computer . Proceedings, Spaceborne Computer Conference, Anaheim, California,
October 30-31, 1962c
Nagy, George, A Survey of Analog Memory Devices . IEEE Transactions on
Electronic Computers, August 1963, Vol. EC-12, No. 4, pp. 388-393-
NCR Technical Brochure, Description of CRAM - Card Random Access
Memory . 1963, NCR 315, MD 315, lOlA.
NCR Sales Brochure, NCR CRAM Card Random Access Memory . 1962, SP 1555-F20QQQ.
Nelson, R. Co, Magnetic Drums and Discs . Instruments and Control Systems,
January 1962, pp. 109-120.
Newhouse, V. I. and Fruin, R. E., A Cryogenic Data Addressed Memory .
Proceedings Spring Joint Computer Conference, San Francisco, Calif.,
May 1-3, 1962, Vol. 21, pp. 89-100.
Oshima, S., Futami, K., and Kamibayashi, T., The Plated-Wire Memory
Matr ix . 1964 Proceedings Intermag Conference, April 1964, pp. 5-1-1 - 5-1-6.
Pearson, R. T., The Development of the Flexible-Disc Magnetic Recorder .
Proceedings of the IRE, January 1961, Vol. 49, No. 1, pp. 164-174.
Petschauer, R. J. and Turnquist, R. D., A nondestructive Readout Film
Memory , Proceedings Western Joint Computer Conference, Los Angeles,
California, May 9-11, 1961, Vol. 19, pp. 411-427.
Pritchard, J. P. and Wald, L. D., Design of a Fully Associative Cryogenic
Data Processor . Proceedings 1964 Intermag Conference, Washington, D. C.
April 1964, pp. 2-5-1 .
Procyk, F. J. and Young, L. H., A High-Speed Card Changeable Permanent
Magnetic Memory--The Inverted Twistor . 1964 Proceedings Intermag Conference,
Washington, D.C., April 1964, pp. 8-6-1.
g
Pohm, A. v., Zingg, R. J. Hoper, J. H. and Stewart, R. M., Analysis of 10
Element Magnetic Film Memories Systems . 1964 Proceedings Intermag Conference
Washington, D.C., April 1964, pp. 5-3-1.
Pohm, A. v., and Mitchell, E. N., Magnetic Film Memories. A Survey . IRE
Trans, of El Computer, Sept. I960, Vol. EC 9, pp. 308-314.
Raffell, J. I., Future Developments in Large Magnetic Film Memories .
Ninth Annual Conference on Magnetism and Magnetic Materials, Atlantic
City, N. J., November 1963.
5-17
Raf fel , J. I., et al, Magnetic Film Memory Design . Proceedings of the
IRE, January I96I, Vol. kS, No. 1, 155-164.
Rajchman, J. A., Computer Memories - Possible Future Developments .
RCA Review, June 1962, Vol. 23, pp. 147-151.
Rash, Kc H., NCR's Card Random Access Memory (CRAM: Informatics Discfile
Symposium, Hollywood, California, March 6-7, 1963 »
RCA Laboratories, Digital Computer Peripheral Memory . First Quarterly
Report, U.S.A.E.R.&D.Lc Contract, No. DA36-039-AMC-03248 (E), Sept. 30, 1963
RCA Laboratories, Digital Computer Peripheral Memory . Second Quarterly
Report, USAER&DL Contract, No. DA 36-039-AMC-03248 (E), October 1, 1963 -
December 31 , 1963.
Randex Storage System, Remington Rand Sales Brochure . I96I, U 2613, Rev. 1.
Renard, A. M., et al, Non-Destructive Readout Magnetic Thin-Film Memory .
1963 Pacific Computer Conference IEEE, Pasadena, Calif., March 15-16, 1963,
pp. 78-95.
Rowland, C. A. and Berge, W. 0., A 300 Nanosecond Search Memory . Proceedings
Fall Joint Computer Conference, Las Vegas, Nevada, November 12-14, 1963,
Vol . 24, pp. 59-66.
Sallo, J. S., Plated Holes - Simplify Memory Design . Electronics, McGraw
Hill Publication, November 1, 1963, Vol. 36, No. 44, pp. 34-36.
Sampson, D. K., The Data Products Discfile Series dp/f5020. Informatics
Inc. Discfile Symposium, Hollywood, California, March 6-7, 1963.
Schick, Thomas, Discfile Sorting. ACM Sort Symposium, Princeton, N.J.,
November 29-30, 1962.
Seeber, R. R. and Lindguist, A. B., Associated Logic for Highly Parallel
Systems . Proceedings, Fall Joint Computer Conference, November 1963.
Shahbender, R., et al. Laminated Ferrite Memory . Fall Joint Computer
Conference, Las Vegas, Nevada, November 12-14, 1963, Vol. 24, pp. 77-90.
Shahbender, R., et al, Microaperature High-Speed Ferrite Memory . Proceedings
Fall Joint Computer Conference, Philadelphia, Pa., December 1962, Vol. 22,
pp. 197-212.
5-18
Stammer John, L. W., An Evaluation of Design and Performance of the
Permanent Magnet Twistor Memory , 196^ Proceedings Intermag Conference,
Washington, D.C., April 1964, pp. 8-4-1 c
Stein, Irving, Generalized Pulse Recording , IEEE Transactions on
Electronic Computers, April 1963, Vol. EC-12, No. 2, pp. 1 77-92.
Taub, D. M., A Short Review of Read Only Memories , Proceedings IEEE
(British), January 1963, Vol. 110, No. 1, pp. 157-166.
Teer, K., Achievements and Problems in Magnetic Recording . Proceedings
1964 Intermag Conference, Washington, D.C., April 1964, pp. 3-1-1 »
Thomas, L. M. and Lawrence, P. D., One-Core-Per-Bi t Operation of Microferrite
Array , RCA Technical Bulletin, October 1963.
Thornley, R.F.M., Brown, A. V. and Speth, A. J., Electron Beam Recording
of Digital Information , IEEE Transactions on Electronic Computers,
February 1964, Vol. EC-13, No. 1 pp. 36-40.
Tuttle, Gc T., How to Quiz A Whole Memory at Once , Electronics, 36, 46,
November 15, 1963, pp. 43-46.
Uc S. Government Research Departments, Information Storage and Retrieval ,
OTS, Deptc Commerce, January 1962, Vol. 37, PP» 29.
Vinal, A. W., The Development of a New Nondestructive Memory Element ,
Proceedings Western Joint Computer Conference, Los Angeles, California,
May 9-11, 1961, Vol. 19, pp. 443-474.
Wheeler, D. J., Read Only Stores for the Control of Computers , Symposium
on Advanced Computer Organization, IFIPS-62 Conference, Munich, August 1962.
Wieselman, Irving L., Stuart-Williams, Raymond, A Multiple Access Discfile ,
Proceedings Fall Joint Computer Conference, Las Vegas, Nevada, November
12-14, 1963, pp. 351-363.
Wiesner, E. P., Memory Has Nondestructive Readout of Standard Ferrite
Cores , EDN, March 1964, pp. 44.
Williams, R. Stewart and Wieselman, File Storage-Existing File-Storage
Systems and the Design of Discfiles . Special Technical Presentation,
Data Products Corp. 1962.
5-19
Weisz, R. S. and Rosenberg, N., Wide Temperature Range Coincident Current
Core Memories , Proceedings Western Joint Computer Conference, Los Angeles,
California, May 9-11, I96I, Vol. 19, pp. 207-214.
Willis, D. G., Platic Neurons as Memory Elements , Lockheed Aircraft,
LMSD-48432-1 (Technical Report), June 1959, (ASTIA No. AD 219-839).
Winters, H., A Noise Cancelling Two-Core-Per-B i t Nondestructive Readout
Technigue Operated in a Flux Rotation Mode . 1964 Proceedings Intermag
Conference, Washington, D.C., April 1964, pp. 8-2-1.
Wolff, M. F., What's New in Computer Memories , Electronics, 36, 45,
November 8, I963, pp. 35-39.
Yovits, Mc C, Large Capacity Memory Technigues for Computing Systems ,
Macmillan, N.Y., 1962.
5-20
5.2.4 Bibliography; Components and Packaging
Abeyta, I., Bargini, F. , and Crosby, D. R. , A Computer Subsystem
Using Ki lomegacyc 1e Subharmonic Oscillators , Proceedings, IRE,
January 1961, Vol. 49, No. 1, pp. 128-135.
Anderson, G. P., and Erickson, R. A., Failure Modes in Integrated
and Partially Integrated Mi croelectr i c Circuits , Proceedings of
Second Annual Symposium on the Physics of Failure in Electronics,
Sept. 25-26, 1963.
Angell, J. B., Information Redundancy and Adaptive Structures , Digest
of Technical Papers, pp. 84-85, 1964 Solid State Circuits Conference,
Philadelphia, Pa., Feb. 19-21, 1964.
Bailey, D. , Integrated Linear Circuits , Electronic Products, June 1964,
pp. 50.
Baker, T. , and Dillon, C, All Magnetic Digital Circuits and Application
Problems, Digest of 1964 International Solid State Circuits Conference,
Philadelphia, Pa., Feb. 1964, pp. 18-19.
Bennion, D. R. , et al, A Bibliographical Sketch of All Magnetic Logic
Schemes , IRE Transactions on Electronic Computers, June 1961, Vol. EC 10,
No. 2, pp. 203-206.
Bremer, J. W. , Superconductive Devices , McGraw-Hill Book Co., New York,
1962.
Bridges, J. M. , Government Needs and Policies in the Age of Microelectronics ,
The Impact of Microelectronics, McGraw-Hill Publishing, N.Y., N.Y., 1963,
pp. 31-40.
Buddenhaga, D. A., Lasers and Their Applications , Society of Automotive
Engineers, Detroit, Michigan, Jan 13-17, 1964.
Carr, W. N. , and Milnes, A. G., Bias Controlled Tunnel-Pair Logic Circuits ,
IRE Transactions on Electronic Computers, Dec. 1962, Vol. EC- 1 1 , No. 6,
pp. 773-779.
Carr, E. Q.. , Flip Chips Easier to Connect, Electronics, Oct. 18, 1963,
pp. 82-84.
Chambers, W. S., Opto-E lectronic Devices in Microelectronics, 16th Annual
NAECON, Dayton, Ohio, May 13, 1964.
5-21
Chesarek, Donald J., Logical Limitations of Gigahertz Circuits ,
Computer Design, Sept. 1963, Vol. 2, No. 8, pp. 16-23.
Comfort, W. T. , A Modified Holland Machine , Proceedings Fall Joint
Computer Conference, Las Vegas, Nev . , Nov. 12-14, 1963, Vol. 24, pp.
481-488.
Connor, J. A., An Inventory of Intrinsic Sources of Chance Failures in
Electronic Parts, 1964 Electronic Components Conference, Wash., D.C.,
8 May 1964.
Control Engineering, Integrated Mi croci rcu i ts and Control: The Coming
Impact , McGraw-Hill Publication, Nov. 1963, pp. 67-71.
Cook, C. R. , and Martin, B. M. , New Semiconductor Networks Reduce System
Complexity , McGraw-Hill Publication, Jan. 10, 1964, Vol. 37, No. 2, pp.
25-29.
Cooperman, M. , 300 mcs Tunnel Diode Logic Circuits , 1963 Pacific Computer
Conference IEEE, Pasadena, Calif., March 15-16, 1963, pp. 166-186.
Grain, N. R. , Nanosecond Circuits and Interconnection Techniques , 1963
Pacific Computer Conference IEEE, Pasadena, California, March 15-16, 1963,
pp. 131-138.
Crane, H. D. , and Van DeRiet, E. K. , Design of an All Magnetic Computing
System: Part |-Circuit Design, IRE Transactions on Electronic Computers,
June 1961, Vol. EC 10, No. 2, pp. 207-220.
Cubert, J. S., and Chow, W. S., Enhancement Tunnel Diode Nanosecond Logic
Ci rcu i ts , 1963 Pacific Computer Conference IEEE, Pasadena, Calif., March
15-16, 1963, pp. 187-193.
Davies, D. C, Seeds, R. B., and Shou, S. H., An Integrated Charge Control
J-K Fl ipFlop , Digest of Technical Papers, pp. 70-71, 1964 Solid State
Circuits Conference, Phi la.. Pa., Feb. 19-21, 1964.
Davis, E. M. , Harding, W. E., Swartz, R, S., and Korning, J. J., Sol id
Logic Technology: Versatile, H igh-Performance Micro-Electronics , IBM
Journal of Research & Development, Vol. 8, No. 2, pp. 102-114.
Deuber, G., Microminiature Circuits Used for Compatible IBM 360 System ,
Electronic News, April 13, 1964, p. 4
Electrochemical Design, Integrated Circuits Seminai — Part I , June 1963,
pp. 70-75.
5-22
Flectronics. Big Computer Goes in Small Package , McG raw H i II Publication,
March 13, 1954, p. 28.
Flectronics. Crystals Route Laser Beam , McGrawHill Publication, Feb. 7,
1964, p. 67.
Electronics, Flat Cable Shrinks Microcircuit Modules , McGraw-Hill
Publication^ Feb, 21, 1964, pp. 44-45.
Flectronics. Integrated Circuits More Reliable , McGraw-Hill Publication,
Feb. 28, 1964, pp. 19.
Electronics, Ki 1 I i ng the Parasitics , April 6, 1964, p. 29.
Flectronics. Microcomputer Comes Off the Line , McGraw-Hill Publication,
Nov I, 1963, Vol. 36, No. 44, pp. 14-15.
Flectronics. Standards Proposed for Flatpacks , McGraw-Hill Publication,
Jan 17, 1964, p. 52-55.
Electronics Magazine and Armour Research Foundation, The Impact of
Microelectronics , Proceedings of Conference on Impact of Microelectronics,
June 26, 27, 1963.
Electronics, 1964: The Year Mi crocj rcu i ts Grew Up , McGraw-Hill Publication,
March 13, 1964, pp. 10-11.
Electro Technology Staff Report, Microelectronic Components: Capability
and Aval labi 1i ty , Oct. 1963, pp. 103-108.
Fairchild Technical Bulletin, Fairchild Epitaxial Micrologic , Oct. 1963,
A-64 Rl.
Falkner, A. H., Magic-An Advanced Computer for Spaceborne Guidance Systems ,
Proceedings, Spaceborne Computer Engineering Conference, Anaheim, Calif.,
Oct. 30-31, 1962, pp. 83-94.
Feldman, C, The Future of Thin Film Active Devices , Electronics,
McGraw-Hill Publication, Jan. 24, 1964, Vol. 37, No. 4, pp. 23-26.
Fow 1 e r , A . B . , Thin Film Circuit Technology: Part lll--Active Thin Film
Devices , IEEE Spectrum, June 1964, Vol. I, No. 6, pp. 102-111.
Gaertner, W. W. , Heizman, C. , Levy, C, and Schuller, M. , Microelectronic ,
Micropower Digital Circuits and Low Level Amplifiers for Space Applications ,
Proceedings Spaceborne Computer Engineering Conference, Anaheim, Calif.,
Oct. 30-31, 1962, pp. 151-166.
5-23
Garibotti, D. J., The Enchanced Micro-Module, A Universal Inter -
connection-Packaging System , 1964 Electronic Components Conference,
Washington, D.C., May 8, 1964.
Garrett, E., and Roby, L. E., Solving Interconnection Problems Resulting
from the Microminiaturization Trend , Proceedings MIL E CON 7, Washington,
D.C., Sept. 10-11, 1963, pp. 354-356.
Garth, E., System Design with MECL Integrated Circuit Logic Blocks .
Motorola Semiconductor Technical Information, Sept. 1963, IC 11.
Genser, M. , Smith, Serett, Failure Modes in Thin Film Circuits , Proceedings
1964 Electronic Components Conference, Washington, D.C. May 8, 1964,
Gerhold, R. A. , Effective Packaging of Micro-Electronics with Micro-Modules ,
U.S. Proceedings MIL E CON 7, Washington, D.C, Sept. 10-11, 1963, pp. 357-360
Gilleo, M. A., and Last, J. T. , Optical Coupling: New Approach to Micro -
Circuit Interconnections, Electronics, McGraw-Hill Publication, Nov. 22, 1963,
pp. 23-27.
Gilligan, T. E., and Roop, D. E., Integration of Nanosecond Emitter -
Coupled Logic , Digest of Technical Papers, pp. 74-75, 1964 Solid State
Circuits Conference, Phi la.. Pa., Feb. 19-21, 1964.
Gottlieb, E. and Giorgis, J., Tunnel Diodes a Four Part Series , Electronics,
June 14, 21, 28, July 5, 1963.
Gould III, E. B., and Wiley, C. A., Realization of the Reliability
Potential for Microelectronics , National Winter Convention on Military
Electronics, Los Angeles, Feb. 5-7, 1964, Vol. Ill, pp. 2-61 - 2-67.
Haggarty, P. E., The Economic Impact of Integrated Circuitry , IEEE
Spectrum, June 1964, Vol. 1, No. 6, pp. 80-82.
Harmon, E. F. , Interconnection of Integrated Circuit Flat Packs in
Autonetics Improved Minuteman Program , Electronic Components Conference,
Washington, D.C, May 8, 1964,
Haun, R, D. , Laser Materials and Devices — A Research Report , Electro-
Technology, Conover-Mast Publication, Sept. 1963, pp. 63-71.
Haynes, John L., Logic Circuits Using Square- Loop Magnetic Devices: A
Survey , IRE Transactions on Electronic Computers, June 1961, Vol. EC-10,
No. 2, pp. 191-202.
Henkel, R. , Integrated Circuit Computers in Production , Electronic News,
Oct. 28, 1963, pp. 42.
5-2i*
Hirson, J. M. , Kaplan, Pollino, Thick Fi Im Hybrids , 1964 Electronic
Components Conference, Washington, D.C., May 8, 1964.
Hodges, D. A., Pederson, D. 0., and Pepper, R. S., A Simple integrated
Realization of a Bistable Circuit , Digest of Technical Papers, Solid
State Circuits Conference, Phi la, Pa., Feb. 19-21, 1964, pp. 72-73,
Hohman, J, W. , Recent Advances in Microelectronics , Proceedings 1963 MIL
E CON 7, Washington, D.C., Sept. 10-11, 1963, pp. 329-332.
Holbrook, E. L. , Pneumatic Log (i-lv). Control Engineering, July, August,
November 1961 and February 1962.
Holloway, J. A., Monolithic Digital Circuits . E D N, Feb. 1964, p. 18.
Honeywell Electronic Data Processing, Mi Idata Study, Quarterly Progress
Report #1 , Aug. 12, 1963, to Nov. 8, 1963, DA-36-039-AMC-03275 (E) ,
3 Dec. 1962,
Hulme, J. R. , Integrated Circuit Design Techniques, Electronic Industries,
April 1963, pp. 112-115.
Integrated Circuits Associates, Integrated Circuits - A Technical Review
and Business Analysis , Results of a Harvard Business School Study, 1963.
Ittner, W. B., The Case for Cryogenics? , Proceedings, 1962 Fall Joint Computer
Conference, Dec. 1962, pp. 229-231.
Jones, W. N. , Cricchi, J. R. , List, W. F. , A Functional Electronic Block
for General Circuit Applications , Seventh MIL E CON, Sept. 9-11, 1963.
Kapany, N. S., Fiber Optics and the Laser , paper presented at New York
Academy of Sciences Conference on the Laser, New York, N.Y,, May 4-5, 1964.
Keonjian, E., Micro Electronics , McGraw-Hill Publishing Co., New York, N.Y.,
1963.
Kl Iby, J. S., Interconnection Techniques for Semi-conductor Networks ,
Proceedings Western Joint Computer Conference, Los Angeles, California,
May 9-11, 1961, Vol. 19, pp. 87-94.
Koster, C, Possible Uses of Lasers in Optical Logic Functions , 1963 Pacific
Computer Conference, IEEE, Pasadena, California, March 15-16, 1963, pp. 54-62.
5-25
Kulcke, W. , et al, A Fast Digi tal- I ndexed Light Deflector . IBM Journal
Jan. 1964, Vol. 8, No. 1, pp. 64-67.
LaFond, C. D. , Billion-Dollar Annual Market is Due to Double by the End
of the Decade . Special report in Missiles and Rockets, June 1, 1964,
pp. 23-58.
Lambert, A., Trends in Logic Circuit Design , Electronics, McGraw-Hill
Publication, Dec. 6, 1963, Vol. 36, No. 49, pp. 33-45.
Lampathakis, K. , et al, New Ultra High Speed Integrated Logic Circuits ,
1963 Pacific Computer Conference IEEE, Pasadena, California, March 15-16,
1963, pp. 8-17.
Langlois, P. J., Tunnel Diodes Boost TRW Speed , Electronics, May 10, 1963,
Vol. 36, No. 19, pp. 50-51.
Lengyel, B. A., Lasers , John Wiley £• Sons, 1964.
Lo, A. W. , Some Thoughts on Digital Components and Circuit Techniques ,
IRE Transactions on Electronic Computers, Sept. 1961, Vol. EC-10, No. 3,
pp. 416-425.
Luedicke, E., and Medwin, A., Microsystem Computer Techniques . Proceedings
Western Joint Computer Conference, Los Angeles, California, May 9-11, 1961,
Vol. 19, pp. 95-110.
Lydon, J., Integrated Circuits Seen Cutting U.S. Costs 7% in Next 10 Years ,
Electronic News, March 25, 1964.
Maclntyre, R. C. , Interconnections of Organizations of Functional Electronic
Blocks , 16th Annual NAECON, Dayton, Ohio, May 13, 1964.
Maiman, T. H., The Laser Field. . .Where We Stand Today , Electronic Industries,
Feb. 1964, pp. 68-74.
McKenzie, A. A., Modern Electronics Packaging . Electronics, McGraw-Hill
Publication, Feb. 7, 1964, pp. 33.
Meyers, P. B. , A Survey of Microsystem Electronics , Proceedings Western
Joint Computer Conference, Los Angeles, Calif. May 9-11, 1961, Vol. 19,
pp. 63-74.
Miller, B. , Thin Film Transistor Research Pressed . Aviation Week and Space
Technology, Feb. 4, 1963, pp. 68-77.
5-26
Mlnnick, R. C, and Short, R. A., Cellular Li near- Input Logic , Final
Report on AF19 (628) -498 , Project 4641, Task 464101, Stanford Research
Institute, February 1964,
Motorola Technical Bulletin, Advanced Fabrication Techniques for Motorola
Integrated Circuits , IC 101, Dec, 1962.
Motorola Technical Bulletin, Motorola Customline Diode-Transistor Logic
Integrated C i rcu i ts , Jan , 1964, #4249,
Nathat, M. I., and Burns, G., Injection Lasers: State of the Art,
Electronics, McGraw-Hill Pbulication, Dec. 6, 1963, Vol. 36, No. 49,
pp, 61-65.
Naymik, D. A., Silicon Mosiac for Integrated Devices , Solid State Circuits
Conference, Philadelphia, Pa., Feb. 19-21, 1964, Digest of Papers, pp, 52-53,
Newell, W, E., The Frustrating Problem of Inductors in Integrated Circuits
Electronics, McGraw-Hill Publication, March 13, 1964, pp, 50-52.
Newhall, E. E., All Magnetic Digital Circuit Fundamentals , Digest of 1964
International Solid State Circuits Conference, Phi la,, Pa., Feb. 1964,
pp. 16-17.
Northrup, J. B., New Interconnection Methods for Micro-Circuits , 1964
Electronic Components Conference, Washington, D.C., May 8, 1964.
Noyce, R. N. , Integrated Circuits in Military Equipment , IEEE Spectrum,
June 1964, Vol. I, No. 6, pp. 71-72.
O'Connell, E. P., and Brauer, J. B., Microelectron ics-Where, Why, and
When , Proceedings National Winter Convention, MIL E CON, Los Angeles,
California, Feb. 5-7, 1964, Vol. Ill, pp. 2-1.
Pacific Semiconductor, Inc., Utilization of New Techniques and Devices in
Integrated Ci rcu I ts . Second Quarterly Report, AF Contract AF33(657) - 1 1 185,
1 Aug. 1963 - 31 August 1963.
Pacifico, E. M. and King, Stanley, Integrated Circuits Shrink a Doppler Radar
System . Electronics, McGraw-Hill Publication, March 23, 1964, pp. 74-79.
Palmer, J. M., Integrated Circuit Development for Digital Communication
Equ ipment . Digest of Technical Papers, Solid State Circuits Conference,
Phi la., Pa., Feb. 19-21, 1964, pp. 76-77.
Parkinson, G., Circuit 95% Integrated in Airborne Digital , Electronic
News, April 20, 1964, p. 30.
5-27
Peck, D. S. , Reliable Systems from Reliable Components . Digest of
Technical Papers, 1964 Solid State Circuits Conference, Phi la., Pa.,
Feb. 19-21, 1964, pp. 80-81.
Peel, M. E., A New Concept for Mi crominature Interconnections . 1964
Electronic Components Conference, Washington, D.C., May 8, 1964.
Peil, W. , et al., UHF Computer Circuits , 1963 Pacific Computer Conference,
IEEE, Pasadena, California, March 15-16, 1963, pp. 163-165.
Phelps, M. , Choice of Logic Forms for Integrated Circuits , Electrical
Design News, Cahners Publishing Co., Jan. 1964,
Philips, A. B., Monolithic Integrated Circuits , IEEE Spectrum, June 1964,
Vol. 1 , No. 6, pp. A-3 - 101 .
Platzek, R. C. and Goodman, H. C. Digital Computer Aspects of Integrated
Circuit Applications , Proceedings, National Winter Convention on Mil.
Electronics, Los Angeles, California, Feb. 5-7, 1964, Vol. Ill, pp.
2-34 - 2-53.
Pleshek, L. J., Criteria for Technical Analysis as Applied to the Selection
of Military Subsystems for Microminiaturization . Proceedings, National
Winter Convention on Military Electronics, Los Angeles, California,
Feb. 5-7, 1964, Vol. Ill, pp. 202-206.
Powers, G., System Speed with Integrated Circuits , E D N, May 1964.
Rice, R. , Systematic Procedures for Digital System Realization from Logic
Design to Production . Digest of Technical Papers, 1964 Solid State Circuit
Conference, Phi la., Pa., Feb. 19-21, 1964, pp. 8-9.
Ridgway, W. C, An Adaptive Logic System with Generalizing Properties ,
Stanford Electronics Laboratories, April, 1962, (ASTIA No. AD 286-914).
Rischall, H., Laser Welding of Microelectronic Interconnections, 1964
Electronic Components Conference, Washington, D.C, May 8, 1964.
Robinson, D. D. , The System Approach to Microminiaturization , Proceedings
National Winter Convention on Military Electronics, Los Angeles, Calif.,
Feb. 5-7, 1964, Vol. Ill, pp. 2-7 - 2-24.
Rosengreen, A., Preliminary Feasibility Study of Neuristor Devices , Final
Technical Report (ASD-TDR- 198) , Stanford Research Institute, Jan. 1963,
(ASTIA No. 402-664) .
5-28
Schlosser, W. , Lascaro, C. , and Key, J. , Pulsed Nuclear Radiation
Effects on Electronic Components £• Materials , Electro- Technology,
May 1963.
Schultz, W. R. , and Brenner, J. H. . Application of Multi-Aperture
Devices in Space Borne Digital Control Equipment , Proceedings MIL
E COM 7, Waslnington, DC, September 10-11, 1963, pp. 425-429.
Schwartz, Lloyd, Microci rcui ts , Mechanization Could Remold Industry :
BLS EDN, April 20, 1964.
Shaw, Donald M. , A Description of the Use of Semiconductor Integrated
Circuits for Artillery Timers and Other Ordinance Material . Nat' I
Winter Convention on Military Electronics, Feb. 5-7, 1964, Vol. Ill,
Los Angeles, pp. 2-25-233.
Sims, R. C. , Becl<, Jr., E. R. , and Kamm, V. C. , A Survey of Tunnel -
Diode Digital Techniques , Proceedings of the IRE, January 1961,
Vol. 49, No. 1 , pp 136-146.
Skoures, Alex B. , Choosing Logic for Microelectronics , Electronics,
McGraw Hill publication, October 4, 1963, Vol. 36, No. 40, pp 23-26.
Snitzer, E. , and Koester, C. J., Some Properties of Fiber Optics
and Lasers , presented at Symposium on Optical Processing of Infor-
mation, Washington, D.C., October 23-24, 1962.
Socolovsky, A. , Diced. Cemented Silicon Mosaic , EDN, May 1964.
Socolovsky, A. , Microelectronics for a Commercial Computer . EDN,
May 1964.
Space Age News, Laser Researchers Seek New Materials ; Transition to
Industrial Uses Progressing , November 1963, pp. 14-18.
Spandorfer, L. M. , Topological Aspects of Microcircuit Interconnec -
tions . 16th Annual NAECON, Dayton, Ohio, May 13, 1964.
Special Report ; Microelectronics . Staff of Missiles & Rockets, Feb. 3,
1964, Vol 14, No. 5, pp. 22-63.
Sturman, J. C. , Micropower Transistor Logic Circuits . NASA TN D-1462
Technical Note, February 1963.
Sturman, J. C. , Recent Developments in Semiconductor Circuits , Confer-
erence on Industrical Applications of New Technology, Atlanta, Georgia,
April 2-3, 1964.
5-29
Suran, J. J. , Use of Circuit Redundancy to Increase System Reliability ,
Digest of Technical Papers, Feb 19, 20, 21, 1964, Solid State Cir-
cuits Conference, Phila., Pa., pp. 82-83.
Telfer, Thomas, Hermetic Packages for Microsystems Electronics , Pro-
ceedings MIL E Con 7, Washington, DC, September 10-11, 1963, pp. 350-353.
Tippett, J. T. , The Status of Optical Logic Elements for Nanosecond
Computer Systems , 1963 Pacific Computer Conference IEEE, March 15-16,
1963, Pasadena, Calif., pp. 47-53.
Uzunoglu, Vasil, Distributed Parameters in Molecular Structures ,
Proceedings MIL E CON 7, Washington D.C. , September, 10-11, 1963,
pp. 341-344.
Vogel , S. and Dulberger, L. H. , Lasers: Devices and Systems - Part I ,
Electronics, October 27, 1961.
Waller, Larry, Advances in Hybrid Integrateds at TRW . Electronic News,
March 25, 1964, p. 29
Weber, Samuel, Optoelectronics' Advance Slows Down . Electronics, Feb.
28, 1964, McGrawHill Publication, pp. 10-11.
Weimer, P. K. , Borkan, H. , Meray-Horvath, L. Shall cross, F. V. ,
An Integrated Thin Film Image Scanner , Digest of Technical Papers,
1964 Solid State Circuits Conference, Phila., Pa., Feb. 19, 20, 21,
1964, pp. 68-69.
Westinghouse Technical Bulletin, Integrated Circuits , March 1964, 91-100.
Whit^, G. R. , Gas Lasers , Electro-Optical Systems, Inc. Pasadena, Calif.
White, G. R. , Review of Laser Applications . 16th Annual NAECON,
May 12, 1964, Dayton, Ohio.
Winder, R. 0. Threshold Logic in Artificial Intelligence, Radio Corpor-
ation of America, Scientific Report No. 6, November 16, 1962, (ASTIA
No. 298-784).
Wolff, M. F. Advances in Microminiaturization , Electronics, February
15, 1963.
Wolff, Michael E. , Thin-Film Transistors Form Scanning Geaerator ,
Electronics, McGrawHill Publication, Feb. 21, 1964, pp. 23-25.
Wolff, Michael F. , Cnmpnl-pr in thft M i rrnr. i rriii f Design Rnnm^
.Electronics, McGraw Hill Publication, March 23, 1964, pp. 100-104.
Yaeger, Don, Micropower Microelectronic Goals Detailed at Symposium ,
EDN April 20, 1964.
5-30
5.2.5 Bibliography; Advanced Usage Techniques
Bauer, W. F. , Computer Design from the Programmer's Viewpoint . Pro-
ceedings, Eastern Joint Computer Conference, December, 1958.
Bauer, W. F. , Why Multi-Computers ?, Datamation, September, 1962.
Bauer, W. F. , and Simmons, Sheldon, The PMR Real-Time Data Handling
System , to be published in Datamation, January 1964.
Bauer, W. F. , and West, G. P., A System for General Purpose Analog -
Digital Computation , ACM Journal , 1957, Vol. 4, pp. 12-17.
Boutwel 1 , E. 0., and Hoskinson, E. A., The Logical Organization of the
PB 440 Microprogrammable Computer , Proceedings Fall Joint Computer
Conference, November, 1963.
Brooks, Jr., F. P., Advanced Computer Organization , Proceedings IFIP-62
Conference, August, 1962, Munich, Germany.
Chapin, G. G. , Organizing and Programming a Shipboard Real-Time Compu -
ter System , Proceedings Fall Joint Computer Conference, November, 1963.
Comfort, W. T. , A Modified Holland Machine . Proceedings Fall Joint Com-
puter Conference, November, 1963.
Dittmann, D. W. , Introduction to Navy Tactical Data Systems , National
Convention on Military Electronics, September 11, 1963, Washington, D.C.
Maguire, Thomas, New Breed of Computer Sought , (AF Project Forecase) ,
Electronics, McGraw Hill Publication, December 20, 1963, pp. 24-25.
Masson, John F. , Next for Navy: Integrated Avionics . Electronics, Feb.
28, 1964, McGraw Hill Publication, pp. 43.
Newe 11, A . , Shaw , J , C . , Si mon , H . A . , Empirical Explorations of the
Logic Theory Machine; A Case Study in Heuristic . Proceedings, Western
Joint Computer Conference, February, 1957.
Preywes, W. S. , and Litivin, S. . The Multi-List Central Processor . Work-
shop on Computer Organization, Baum and Knapp, Editors, Cleaver-Hume
Press, London, 1963.
Slotnick, D. L. , Borck, W. C. , and McReynolds, R. C. , The Soloman Com-
puter , Proceedings Fall Joint Computer Conference, December, 1962.
5-3
Bushor, W. E. , The Perceptron - An Experiment in Learning , Electronics,
Jully 22, 1960, 33, pp. 56-59.
Armer, P., Attitudes Toward Intelligent Machines . The Rand Corp., Sept.
30, 1960, Santa Monica, P-2114.
Computing News 66, 705 Indexes Dead Sea Scrolls , April 15, 1958.
Holmes, W. S. , Leland, H. R. , and Richmon, G. E. , Design of a Photo
Interpretation Automaton , Proceedings, Dec. 1962, FJCC.
Mathis, S. J., Jr., Sass, M.A. , and Wilcox, R. H. , Heuristic Programs .
Fact, Fad or Futil i ty ? , Proceedings MIL E CON 7, Washington D. C. ,
Sept. 9, 10, 11, 1963, pp. 177-181.
Moles, A.A. , Principles d' Incerti tude de la Perception et Machines Philo-
sophigues , Cybernetics, 2, 1, 1959, pp. 51-57.
Uhr, L. , and Vossler, C. , A Pattern Recognition Program that Generates ,
Evaluates, and Adjusts its Own Operators , Proceedings, WJCC, 1961, Los
Angeles.
Wooldridge, D. , The Machinery of the Brain , McGraw-Hill Book Co., 1963,
New York.
5-32
5.2.6 Bibliography: Machine System Organization
Amdahl, Lowell, Microprogramming and Stored Logic , Datamation, Feb., 1964
pp. 24-26.
Beck, L. and Keeler, F. , The C-8401 , Datamation, Feb., 1964, pp. 33-35.
Blankenbaker , J.V. , Logically Microprogrammed Computers , Trans. P.G.E.C.,
Vol. EC-7, June 1958, pp. 103-109.
Boutwell, E.O. , "The PB-400 Computer ," Datamation, Feb., 1964, pp. 30-32.
Boutwel 1 , E. and E. Hoskinson, The Logical Organization of the PB 440
Microprogrammable Computer , Proc. F. J.C.C. , Nov., 1963, p. 201-213.
Burks, Arthur W. , Goldstine, Herman H. , and von Neumann, John» Prel im -
i nary Discussion on the Design of an Electronic Computing Instrument ;
institute for Advanced Study, June 1946.
Forest, Robert B. , System/360's Initial Impact , Datamation, May, 1964,
pp. 68-71.
Glantz, H.T. , A Note on Microprogramming , Journal of the Associat ion
for Computing Machinery , Apr i 1 , 1956.
Grassel 1 i , A. , The Design of Program-Modifiable Micro-Programmed
Control Units , IRE Transactions on Electronic Computers, vol. EC-11,
no. 3, June 1962, pp. 336-339.
Hill, Richard H. , Stored Logic Programming and Applications . Datamation,
Feb. , 1964, pp. 36-39
Hill, Richard H. , Stored Logic Revisited , Los Angeles Chapter of ACM,
Dec. 6, 1961.
Kampe, T.W. , The Design of a General -Purpose Microprogram-Controlled
Computer with Elementary Structure , IRE Transactions on Electronic
Computers, vol. EC-9, no. 2, June 1960, pp. 208-213.
Mercer, Robert J., M i cro- Programm i ng . Journal of the Associat ion for
Computing Machinery, April 1957.
McGee, W. C. , The TRW- 133 Computer , Datamation, Feb., 1964, pp. 27-29.
5-33
P^ige, L. J., and Tompkins, C.B. , Scamp Postscript No. 1, Systematic
Generation of Permutations on an Automatic Computer and an Appi tcation
to a Problem Concern ibg Finite Groups ; National Bureau of Standards,
Jan. , 1953.
Patrlcl<, R. L. , A Customized Computer , Datamation, May - June 1960.
Semarne, H. M. , Porter, R. E. , A Stored Logic Computer , Datamation,
May, 1961.
Wilkes, M. V. , M i c r op r og ramm i n g , Proc. EJC Dec. 3-5, 1958, pp. 18-20.
Wilkes, M. V., and Stringer, J. B. , Micro-programming and the Design of
the Control Circuits in an Electronic Digital Computer , Proceedings
of the Cambridge Philosophical Society, April, 1953.
Wilkes, M. V. , The Best Way to Design an Automatic Calculating Machine ,
Manchester University Computer Inaugural Conference, Proceeding^, July 1951
5-3^
5.2.7 BIbl I'oqraph; Programming
Amdahl, Lowell D. , Microprogramming and Stored Logic . Datamation,
Feb. 1964, Vol. 10 No. 2.
Anderson, J. P., Hoffman, S.A. , Shifman, Joseph, Williams, R. J. ,
D-825 A Multiple Computer System for Command and Control , AFIPS Conf.
Proc. , Vol. 22 1962, FJCC.
Aoki , M, Estvin, G. , Mandel 1 , R. , A Probabal istic Analysis of Computing
Load Assignment in a Multi-Processor Computer System , AFIPS Conf. Proc.
Vol. 24, 1963 FJCC.
Armer, P., Attitudes Toward Intelligent Machines , The RAND Corporation,
Santa Monica, California, P-2114, Sept. 30, 1960.
Austin, Kenneth C. , Scientific Computing , Datamation, June 1964,
Vol. 10, No. 1.
Backus, J.W. , et. al.. Report on the Algorithmic Language ALG0L60 ,
Communications of ACM, 1960, Vol. 3. .
Bagley, P. R. , Improving Problems Oriented Language By Stratifying
I t . Computer Journal, Oct., 1961, Vol. 4, No. 3.
Barclay, A. G. , The Achilles Heel of Data Processing , Proceedings
Computers and Data Processing Society of Canada, June, 1960.
Barton, R. S. , A New Approach To the Functional Design of a Digital
Computer . Proceedings of Western Joint Computer Conference. Los
Angeles, Calif., May 9-11, 1961.
Bauer, W. F. , Frank, W.L., DODDAC - An Integrated System For Data
Processing, Interrogation, and Display , AFIPS Conf. Proc, 1961,
Vol. 20, EJCC.
Bemer, R.W. , Survey of Modern Programming Techniques . Computer Bul-
letin, Mar., 1961, Vol. 4, No. 4.
Benington, H..D., Everett, R.R. , and Zvalset, C. A., SAGE - A 0ata
Processing System for Air Defense, Proceedings Eastern Joint Computer
Conference, Dec. , 1957.
Bissell, S. Edward. Measuring Progranmer's Effectiveness , Data Pro-
cessing, Aug., 1960, Vol. 2, No. 7.
Blatt, John M. .Ye Indiscreet Monitor. Comm. ACM, Sept., 1963, Vol 6, No 9.
5-35
Blumanthal, Sherman C. , On Line Processing . Datamation, June, 1961,
Vol. 7, No. 6.
Bottenbruch, H. , Structure and Use of ALG0L60 . Journal of ACM, April,
1962, Vol.9, No. 2.
Boucher, H. , Organisation et Fonctionnement des Machines Ar i thmetiques ,
1960, Masson et Cie., Paris.
Boyd, A.G. , A General Approach to Information Systems Design , Control
Engineering, Aug. 1962, Vol.9, No 8.
Brachman, R. J., Factory to Foxhole - Army Maintenance System "MAIDS ."
Proc. 1964 National Winter Conv. on Mil. Elec, Vol 3.
Breslow, Donald H. , Built-in Test System for Automatic Fault Detec -
tion Design Approach to Checkout of Complex Systems , Electronics,
June 17, 1960, Vol 33, No. 25.
Brooks, F.P.Jr., Blaauw, G.A. , Bucholz, W. . Processing Data in Bits
and Pieces , IRE Transactions on Electronic Computers, ECS, June,
1959, No. 2.
Brown, J. C. , Loglan , Scientific American, June, 1960.
Bush, R. R. , Estes, W.K. , Studies in Mathematical Learning Theory .
Stanford Univ., Stanford, Calif. 1959^
Campbel 1 , J. G. , Systems Implications of New Memory Developments .
AFIPS Conf. Proc, 1963, FJCC, Vol. 24.
Campsie, J.A. , Advanced Management in Data Processing . Jour. Data
Mgmt. June, 1963, Vol. 1, No. 1 .
Carlson, Walter M. , Computers - The Key to Total Systems Control :
An Industrial Viewpoint . Comm. of ACM , March, 1962, Vol.5, No. 3.
Carr, J.W. ,111, Programming and Coding. Part B of Handbook of Auto -
mation, Computation, and Control , 1959, Wiley, Vol. 2.
Carr, J.W. ,111, Recursive Suscripting Compilers and List-type
Memories . Comm. ACM, 1959, Vol. 2, No. 2.
Chomsky, N. , On Certain Formal Properties of Grammars , Information
and Control, June, 1959, Vol. 2.
5-36
Cllppinger, R. F. , FACT - A Business Compiler: Description and Com -
parison with COBOL and Commercial Translator .
Codd, E.F. , Multiproqram Scheduling . Comm. ACM, June, 1960, Vol. 3, No. 6.
Coffman, E.F. Jr., Schwartz, J.I, Weissman, C. , A General -Purpose Time-
Shaping System , AFIPS Conf. Proc. , 1964, FJCC, Vol. 25.
Coil, E.A. , A Mul tiaddressable Random Access File System , IRE Wescon
Convention Report, Part I, Aug 23-26, 1960.
Collins, George 0. Jr., Experience in Automatic Storage Allocation ,
Comm. ACM, Oct. 1961, Vol.4, No. 10.
Comfort, W.T. , A Modified Holland Machine . AFIPS Conf. Proc, 1963,
FJCC, Vol. 24.
Communications and Electronics, 1912-1962: Human Factors , Proceedings
of IRE, May, 1962, Vol.50, No. 5.
Conway, R.W. , and Maxwell, W.L., CORC - The Cornell Computing Language ,
Comm. ACM, June 1963, Vol.6, No. 6.
Corbato, F. J. , The Compatible Time-Sharing System: A Programmers
Guide , MIT Press, Cambridge, Mass., 1963.
Coulson, John E. (Ed.), Programmed Learning and Computer Based In -
struction , Proc. Conf. Appl. Digital Computers Automated Instruction,
Oct., 1961, John WMey and Sons, New York, 1962.
A Critical Appraisal of COBOL . Computer Bulletin, Mar, 1961, Vol 4, No. 4.
Daniels, A.E. , Some Problems Associated With Large ProgFamming Efforts .
AFIPS Conf. Proc, 1964, Vol.25, FJCC.
Day, R.F. , and Hobbs, C.A. . A Real Time Digital Computer for Radar
Control and Data Processing , Proceedings 6th National Convention
Military Electronics, June 1962, Wash. ,D. C. , (Ava i 1 . from IRE).
Dert, J. J. , and Luke, R. C. , Semi -Automatic Allocation of Data Storage
for PACT I . J. ACM, 1956, Vol. 3, No. 4.
Dijkstra, E.W. , Recursive Programming , Numer. Math., 1960, Vol.2, No. 5.
5-37
Dijkstra, E.W. . Some Meditations on Advanced Programming . Information
Processing 62, Proc. of IFIP Congress 62, North Holland Pub. Co.
Amsterdam.
Dilley, D.R. , Information Retrieval As a Control lership Tool , The
Controller, April, 1961, Vol. 29, No. 4.
Doyle, R.H. , Meyer, R.A. , Bedowitz, R. P. , Automatic Failure Recovery
In a Digital Data Processing System , IBM Journal of Research, Jan. 1959.
Dunn, T.M. , Morrissey, J.H. , Keller, J.M. , Strum, E.C. , Yang, G.H.,
Remote Computing: An Experimental System Part I: External Specifications .
Part 2: Internal Design . API PS Conf. Proc, 1964, Vol. 25, FJCC.
Eckman, Donald B. , Systems: Research and Design . Wiley and Sons,
1961 , New York, N. Y.
Edwards, N.P. , On the Evaluation of the Cost Effectiveness of Command
and Control Systems , AFIPS Conf. Proc, 1954, Vol, 25, :|JvCC.
Ellis, D.O. , and Ludwig, T. J. , Systems Philosophy . Prentice-Hall,
Englewood Cliffs, N. J., 1962.
Ellis, W. , Justus , G. R. , and Bel 1 , W. D. , Systems Talk Through Common -
Language Pool , Control Engineering, Feb., 1961, Vol. 8, No. 2.
Estrin, G. , Fuller, R.H. Some Appl i cat ions for Content-Addressable
Memories . AFIPS, Conf. Proc, 1963, Vol. 24, FJCC.
Fair, R.R. Programming Control by Project Schedule . Datamation, Feb.,
1963, Vol. 9, No. 1.
Farr, Leonard, and Nanus, Burt, Cost Aspects of Computer Programming
For Command and Control , Proc. National Winter Conv. on Mil. Elec,
1964, Vol. 3.
Ferguson, H. Earl, Berner, Elizabeth, Debugging Systems at the Source
Language Level . Comm. ACM, Aug, 1963, Vol. 6, No. 8.
Floyd, R.W. , Kallick, B. , Moore, C»J. and Schwartz, E.S., Advanced
Studies of Computer Programming , ARF Project El 21 Armour Research
Foundation, 1961, Chicago, Illinois.
Floyd, R.W. , A Descriptive Language For Symbol Manipulation . Journal
of ACM, Oct. , 1961, Vol. 8, No. 4.
5-38
Frank, W.L., Gardner, W. H. , Stock, G.L. Programming On-Line Systems .
Datamation, May and June, 1963, Vol. 9, Nos. 5 and 6.
Freed, A. M. . Measuring Human interaction in Man-Machine Systems , IRE
Wescon Convention Record, Part 4, Aug, 1960.
Gainen, Leon, A Simulation Model for Data Systems Analysis . AFIPS
Conf. Proc. , 1961, Vol. 20, EJCC.
Galler, Bernard, A., The Language of Computers . McGraw Hill Book Co.,
New York, 1962.
Gass, S. I., et al. Project Mercury Real-Time Computational and
Data- Flow System . AFIPS Conf. Proc, 1961, Vol. 20, EJCC.
Gauss, E. J. , A Comparison of Machine Organizations by Their Perfor -
mance of the Iterative Solution of Linear Equations , Journal ACM,
Oct. 1959, Vol. 6, No. 4.
Gelerntner, H. , Hansen, J.R. , and Loveland, D.W. , Empirical Explora -
tions of the Geometry Theorem Machine, Proc. WJCC, San Francisco,
Cal if. , May 3t5, 1960,
Gill, S. , Current Theory and Practice of Automatic Programming .
Computer Journal 2,3, October, 1959, 110-114.
Gill, S. , The New Intellectuals ?, Computer Bulletin, Sept. 1961,
Vol. 5, No. 2.
Goode, Harry H. , and Machol , Robert E. , System Engineering . McGraw
Hill Co. , Inc. , 1957.
Goodman, Richard (ED), Automatic Programming , Pergamon Press, Oxford,
England, 1961, Vols. 1 and 2.
Gordon, Geoffrey, A General Purpose Systems Simulation Program , AFIPS
Conf. Proc, 1961, EJCC, Vol. 20.
Gorn, S. , Standardized Programming Methods and Universal Coding .
Journal ACM, 1957, Vol. 4, No. 3.
Gottlieb, C.C. Software Problems , Proceedings Third Conference of
Computer Data Processing Society of Canada, June, 1962, Univ. of Toronto
Press, Toronto, Ontario, Canada.
Grabbe, E.M. , Ramo, S. , Wooldridge, D.E. Handbook of Automation. Compu -
tation, and Control . John Wiley & Sons, New York, 1959, Vol. 2.
5-39
Green, Julien, Symbol Manipulation In XTRAN , Comm. ACM, April, 1960,
Vol. 3, No. 4.
Greene, P. H. , A Suggested Model for Information Representation in a
Computer That Receives, Learns, and Reasons , Proceedings, WJCC, May
3-5, 1960, San Francisco, Calif.
Gurk, H.M. , and Minker, Jack, The Design and Simulation of an Informa -
tion Processing System , Journ. ACM, April, 1961, Vol. 8, No. 2.
Haibt, Lois M. , A Program to Draw Multilevel Flowcharts , Proc. Western
Joint Computer Conf. , March 3-5, 1959.
Hales, A., How to Break the Programming Bottleneck , Data Contr. Aug, 1963,
Vol. 1 , No. 8.
Hal pern, Mark. A Programming System for Command and Control Application .
Proc. 1964 National Winter Conv. on Mil. Elec. , Vol. 3.
Hanssman, F. , Operations Research in Production and Inventory Control .
John Wiley and Sons, Inc., New York, N. Y. 1962.
Head, R. V. , The Programming Gap in Real-Time Systems . Datamation,
Feb. , 1963, Vol. 9, No. 2.
Head, R. V. , Real-Time Programming Specifications. Conm. ACM, July, 1963,
Vol. 6, No. 7.
Heller, J., Seguencinq Aspects of Multiprogramming , Jour. ACM, July,
1961 , Vol. 8, No. 3.
Herman, D. J. The Use of a Computer to Evaluate Computers , AFIPS Conf.
Proc. 1964, SJCC, Vol. 25.
Heskin, Joseph, The Saturn Automatic Checkout System . EJCC Proceedings,
Wash., D.C. , Dec. 1961, Macmillan Co. N.Y.
Hill, R.H. , Stored Logic Programming and Application , Datamation, Feb.
1964, Vol. 10, No. 2.
Hill, W. H. , Electronic Information Systems in Navy Management , Navy
Management Review, Jan. 1959.
Hodskins, J.A. , Machine Utilization Measurement , Journal of Machine Ac-
counting, Dec, 1961, Vol. 12, No. 12.
5-40
Holdiman, T.A. , Management Techniques for Real-Time Computer Programming .
Journal of ACM, July, 1962, Vol. 9, No. 3.
Holland, H.C. , Selecting and Training People for Systems Modernization .
Electronic Data Processing Conference, May 19-20, 1960, Vol. 18, No. 5
(Nov. 1960).
Hoi land , John H. , Outline For a Logical Theory of Adaptive Systems ,
Journal of ACM, July, 1962, Vol. 9, No. 3.
Holland, J., Universal Computer Capable of Executing An Arbitrary
Number of Sub-Programs Simultaneously . Proceedings Eastern Joint Com-
puter Conference, 1959.
Hoi t, A.W. , Program Organization and Record Keeping For Dynamic Storage
A 1 1 oca t i on , Information Processing 62, Proc. of IFIPS Congress 62, North
Holland Pub. Co., Amsterdam.
Hosier, W.A. , Pitfalls and Safeguards in Real-Time Digital Systems With
Emphasis on Programming . IRE Transactions of Engineering Management,
June, 1961 , Vol. EM-8, No. 2.
Howarth, P. J., Jones, B. D. , and Wyld, M.T. , The Atlas Schedul ing
System . Computer Journal, Oct, 1962, Vol. 5, No. 3.
Huskey, Harry D, Halstead, M.H. , and McArthur, R. , NELIAC - Dialect of
ALGOL . Communications of ACM, August, I960, Vol. 3, No. 8.
Israel, David R. , Simulation Techniques for the Test and Evaluation of
Real-Time Computer Programs . Jour. ACM, July, 1957, Vol. 4, No. 3.
Jacoby, I., and Layton, H. , Automation of Program Debugging , Prepoints
of papers presented at the 16th National Meeting of the ACM, Sept.
5-8, 1961, Los Angeles, Calif.
Jeenel , Joachim, Programming for Digital Computers . McGraw Hill Book
Co. , Inc. , 1959, N. Y.
Joachim, Gertrude S. , Memory Efficiency . Jour. ACM, April, 1959, Vol. 6,
No. 2.
Joslin, E.O. , Cost-Value Technigue for Evaluation of Computer System
Proposals , AFIPS Conf. Proc, 1964, Vol. 25, SJCC.
Kaplan, A. , A Search Memory Subsystem For a General Purpose Computer ,
AFIPS Conf. Proc, 1963, Vol. 24, FJCC.
Katz, J.H. , McGee, W.C. , Sears, R.E. . An Experiment In Non-Procedural
Programming . AFIPS Conf. Proc, 1963, Vol. 24, FJCC.
5-41
Kavanagh, T. F. , TABSOL - The Language of Decision Making . Computers
and Automation, Sept., 1961, Vol, 10, No. 9.
Kelburn, T. , Payne, P.B. , Howarth, D.J. . The Atlas Supervisor , AFIPS
Conf. Proc. , 1961, Vol. 20, EJCC.
Keller, Arnold E. , Crisis in Machine Accounting . Management Business
Automation, June, 1961, Vol. 5, No. 6.
Kelley, J.E., Jr., Technigues For Storage Al 1 ocat ion Algorithms . Comm.
of ACM, Oct. 1961, Vol. 4, No. 10.
Kincaid, W. H. , and Simpson, C.H., Use the Editors You Havei . Data
Processing, Aug. 1961, Vol. 3. No. 8.
Klerer, Melvin, Problems in Scientific User Relations , Datamation,
April , 1963, Vol. 9, No. 4.
Koomanoff, F.A. , and Pritsker, A.A.B. , Railroading As a System Con -
cept , Ba telle Tech. Rev., March, 1962, Vol. 11, No. 3.
Lee, Fred, An Automatic Self Checking and Fault Locating Method . IRE
Transactions, Oct., 1962, Vol. EC-ll,No. 5.
Leonard, G.F. , The CL-1 Programming System User's Manual . Technical
Operations Inc. , Jan. 1961 , Burl ington, Mass.
Licklider, J.C.R. , Interaction Between Artificial Intelligence. Mili -
tary Intelligence, and Command and Control , First Congress Information
System Sciences Session, 8 Nov. 1962, Mitre Corp., Bedford, Mass.
Licklider, J.C.R. , Clark, Weldon E. , On-Line Man-Computer Communication ,
AFIPS Conf. Proc, 1962, Vol 21, SJCC.
LISPI - Programmers Manual , Computation Center and Research Laboratories
of Electronics, MIT, Cambridge, Mass. 1960.
Lombard!, Lionel! i. Mathematical Structure of Non-Arithmetic Data
Processing Procedures , Journal of ACM, Jan, 1962, Vol. 9, No. 1.
Lombard!, Lionelli, Non-Procedural Data System Languages , Preprints
of Papers Presented at the 16th National Meeting of the ACM, Sept. 5-8,
1961.
Lombard!, Lionelli. Theory of Files , Conf. Proc, 1960, Vol. 18, EJCC.
5-42
Lucking, J.R. , and O'Neil, J.B. , The Time-Sharing Facilities of the
KDF9 Computer .
Luzzano, V. (Ed) , Systems and Procedures: A Handbook For Business and
Industry , Prentice-Hall, 1959, Englewood Cliffs, N.J.
McCarthy, J., A Basis For a Ma theme tical Theory of Computation , Proceedings
Western Joint Computer Conference, May, 1961, Los Angeles, Calif.
McCarthy, J., HSP/i ; Progranmers Manual , MfT Computation Center and
Research Laboratory of Electronics, March I, 1960.
McCarthy, John, Recursive Functions of Symbolic Expressions and Their
Computation By Machine . Comm. ACM, April, 1960, Vol. 3, No. 4,
McCracken, Daniel D. , Weiss, Harold, Lee, Tsai-Hwa, Progranvninq Busi -
ness Computers , John Wiley and Sons, Inc., 1959, N.Y.
Meacham, Alan D. , and Thompson, Van B. , Total Systems , American Data
Processing, 1962.
Mercer, R. J. , M i croprogramm i ng , Jour. ACM, April, 1957, Vol. 4, No. 2.
Miller, A.E., and Goldman, M. , Organization and Program of the BMEWS
Checkout Data Processor . Proceedings of the Eastern Joint Computer.
Conference, Dec. 13-15, 1960, Vol. 18.
Minsky, M. , Steps Toward Artificial Intelligence . Proceedings IRE,
June, 1961 , Vol. 49, No. 1.
Moshman, Jack, Johnson, Jacob, and Larsen, Madalyn, RAMPS - a Tech -
nique For Resource Allocation and Multi-Project Scheduling , AFIPS
Conf. Proc. , 1963, Vol . 23 SJCC.
Nanus, B, and Farr, L. , Some Cost Contributers To Large-Seal e Programs ,
AFIPS Conf. Proc, 1964, Vol. 25, SJCC.
Nelson, R.A. , How To Write Effective Machine Room Procedures . Data
Processing, July, 1961, Vol. 3, No. 7.
Newell and Tonge, An Introduction To Information/ Processing Language V .
Comm. ACM, April, 1960, Vol. 4, No. 4.
Opler, Aschen, Testing Programming Aptitude , Datamation, Oct. 1953,
Vol. 9, No. 10.
5-43
Orchard-Hayes, William, Another Perspective On Computer Languages . Com-
puters and Data Processing, Jan., 1964, Vol. 1, No. 1.
Perkins, R, and McGee, W.C. , Programmed Control of Mul ti -Computer Sys -
tems , Proceedings IFIP Congress 62, Munich, 1962, North Holland Pub.
Co. , Amsterdam.
Perils, A. J. , and Thornton, C. , Symbol Manipulation by Threaded Lists ,
Comm. ACM, 1960, Vol. 3, No. 4.
Plugge, W.ft, , and Perry, M. N. , Amer ican Ai rl ines ' SABRE Electronic
Reservations System , Proceedings Western Joint Computer Conference,
May 9-11 , 1961.
Rosenthal, S. , Analytical Technigue For Automatic Data Processing
Eguipment Acquisition , AFIPS Conf. Proc. , 1964, \loh. 25, SJCC.
Pollock, Solomon L, Codasyl , Cobol , and Detab-X , Datamation, Feb., 1963,
Vol. 9, No. 2.
Project Mercury Real Time Computational and Data Flow System , Pro-
ceedings Eastern Joint Computer Conference, Dec, 1961, Washington,
D.C. , Macmillan Co. N.Y.
Ream, Norman J., On-L ine Management Information . Datamation, March
and April, 1954, Vol. 10, No. 3 and No. 4.
Redmund , G.H., Mulvihill, D.E., The Use of a Binary Computer For Data
Process ing . Proc. Eastern Joint Comput. Conf., Dec. 13-15, 1960, Vol. 18.
Richardson, L.E. , The Electronic Reservations System for Trans-Canada
Ai r L ines , Proc. Computers and Data Processing Society of Canada,
June, 1960.
Ridgway, A.O. , An Automated Technique For Conducting a Total System
Study , Proc. Eastern Joint Computer Conf. Washington, D.C., Dec. 1961,
Macmi 1 Ian Co. , N. Y.
Riordan, John, Stochastic Service Systems , John Wiley and Sons, Inc.
New York, N.Y. , 1962.
Ronayne, M. F. , The Personnel Side of Automatic Data Processing . Public
Personnel Review, Oct., I960, Vol. 21, No. 4.
Rosen, Saul, A Multi-Language System For Command and Control , Datamation,
Feb. , 1963, Vol. 9, No. 2.
5-44
Rosene, A.F. , Program Design to Achieve Maximum Utilization In a Rea l-
Time Computing System , Proc. Western Joint Computer Conf. , 1959.
Rossheim, Robert, J. , Report On Proposed American Standard Flowchart
Symbols For Information Processing , Comm. ACM, Oct., 1963, Vol. 6,
No. 10.
Rowan, T. C. , Psychological Tests and Selection of Computer Program -
mers , Jour. ACM, Vol. 4, No. 3, 348-353.
Satin , Journal of Air Traffic Control, July, I960. Vol. 3, No. I.
Saxon, James, A. Programmer Training: A Workable Approach , Datamation,
Dec. , 1963, Vol. 9, No. 12.
Scheinberg, Stephen, Note On the Boolean Properties of Context Free
Languages , Information Cont. Dec, I960, Vol. 3, No. 4.
Schlesinger, R.J. , Abbey, K, Erhorn, R. W. , Friedenthal, K. J. , and
Logue, S.H., Principles of Electronic Warfare , Prentice-Hall, Engle-
wood CI iffs, N.J. , 1961.
Scott, A.E,, Automatic Preparation of Flow Chart Listings , Jour, of
the ACM, Jan. 1958.
Seeber. , R.R. , Lindquist, A.B. , Associative Logic For Highly Paral-
lel Systems . AFIPS Conf. Proc. 1963, Vol. 24, FJCC.
Shafritz, A.B. , Miller, A.E., Rose, K. , Multilevel Programming For
a Real-Time System , AFIPS Conf. Proc, 1961, Vol. 20, EJCC.
Shaw, C.J. , JOVIAL - A Programming Language For Real-Time Command
Systems , Annual review in Automatic Programming, Pergamon Press, New
York, 1963, Vol. 3.
Shaw, C. J. , More Instructions Less Work , Datamation, June, 1964,
Vol. 10, No. 6.
Shaw, C. J. , A Programmers Introduction to Basic JOVIAL , System Develop-
ment Corp.," Aug. 7, 1961, Santa Monica, Calif., TM629.
Shaw, C.J. , A Programmers Look at JOVIAL, in an ALGOL Perspective .
Datamation, Oct. 1961, Vol. 7, No. 10.
5-45
Shaw, C. J. , The JOVIAL Manual Part 2 Revision 1. Th e JOVIAL Grammar
and L exicon , System Development Corp., June 9, 1961, Santa Monica, Calif.
TM 555/002/01.
Shaw C. J., JOVIAL , Datamation, June 1961, Vol. 7, No. 6.
Shooman, W. , Parallel Computing With Vertical Da ta, Proc. Eastern
Joint Computer Conference, Dec. 13-15, 1961, Vol. 18.
Shubik, Martin, Approaches to the Study of Decision-Making Relative to the
Fi rm . Journal of Business, University of Chicago, Apr. 1961, Vol. 34,
No. 2.
Simon, Herbert A., and Newell, Allen, Computer Simulation of Human
Thinking and Problem Solving , Datamation, June, 1961, Vol. 7, No. 6,
July, 1961 , Vol. 7, No. 7.
Simon, Herbert, The Hueristic Compiler , The Rand Corp., USAF Project
Rand, 1963.
Slotnik, Daniel, L. , Borck, Carl, W. , McReynolds, Robert, C. , The
Solomon Computer , API PS Conf. Proc, 1962, Vol. 22, FJCC.
Squire, J.S., Palais, S.M. , Physical and Logical Design of a Highly
Parallel Computer , AFIPS Conf. Proc, 1963, Vol. 23, SJCC.
Stewart, W.E., and Crnkovich, J.E., Program Change Procedures , Data-
mation, June, 1964, Vol. 10, No. 6.
Survey of Programming Languages and Processors , Comm. ACM, March, 1963,
Vol. 6, No. 3.
Tatham, Laura, All the Eggs in One Basket at West i nghouse . Data Contr.
Aug, 1963, Vol. 1 , No. 8.
Thompson, F. B. , Fractionizat ion of the Military Co<at(&xt , AFIPS Conf.
Proc , 1964, Vol. 25, SJCC.
Thompson, R. V. , Wilkinson, J. A. , The D825 Automatic Operating and
Schedul ing Program , AFIPS Conf. Proc, 1963, Vol. 23, SJCC.
Thompson, Van B. , PERT, Pro and Con About This Technique , Data Process-
ing, Oct. 1961 , Vol . 3, No. 10.
Thompson, Van B. , A Training Course in Data Processing , Data Processing
Vol. 2, No. 3.
5-46
Tillitt, Harley, Computer Programming For Young Students , Journal ACM
Oct, 1958, Vol. 5, No. 4.
Tonge, Fred M. , Summary of a Hueristic Line Balancing Procedure , Manage-
ment Science, Oct, I960, Vol. 7, No. 1.
Underhill, L.H., The Growth of Complexity Of a General Purpose Program .
Computer Journal, Apr., 1963, Vol. 6, No. 1.
Vazsonyi, A., An On-Line Management System Using English Language . Proc.
WJCC, May, 1961.
Wa 1 1 a ce , Edwa r d , L . Management influence on the Design of Data Pro -
cessing Systems , Harvard Business School, 1961, Boston, Mass.
Wegstein and Youden, W.W. , A Siring Language For Symbol Manipulation
Based on ALG0L60 , Communications ACM, Jan. 1962, Vol. 5, No. 1,
Wei Is, M. B. . MADCAP: A Scientific Compu te r For a Displayed Formula
Textbook Language , Communications of ACM, 1961, Vol. 4.
Wier, J. M. , Digital Data Communication Technigues , Proceedings of
IRE, Jan. , 1961 , Vol. 49, No. 1.
Wilkes, M.V. , M i croprogrammi ng , Proc.EJCC, Dec. 3-5, 1958.
Wilkenson, M. , The JOVIAL Checker, an Automatic Checkout System For
Higher-Level Language Programs , Proceedings Western Joint Computer
Conference, May 9-11, 1961, Los Angeles, Calif.
WIzenbaum, J. , Symmetric List Processor , Comm. ACM, Sept. 1963,
Vol. 6, No. 9.
Young, John W. Jr. , and Kent, Henry, K. , Abstract Formulation of Data
Processing Problems . Jour, of Industrial Engineering, Nov. -Dec. 1958,
Yngve, V.H. , A Model and An Hypothesis For Language Structure , Proc.
American Philosophical Society, Oct., 1960, Vol. 104, No. 5.
5-47
5.3 METHODOLOGY
5.3.1 Computers and Hard Science
5.3 .1 . 1 Technology
Eldridge, F. R. , The Effectiveness of Command Control in Strategic
Operations for the Mid-Sixties , RAND, RM-2152-PR, Oct. 1962 (SECRET)
Franken, P., High-Energy Lasers . Internat. Scl. and Technology,
Oct. 1962.
Hackforth, H. L. , Infrared Radiation, McGraw-Hill, New Yor, 1960
Kahn, H., On Thermonuclear War , Princeton Univ. Press, Princeton, 1960
Kittell, C, Introduction to Solid-State Physics , Wiley, New Yor, 1956,
Kroger, M. G., Computers in Command and Control , Inst, for Defense
Analysis, TR 61-12, Nov. 1961.
(Lasers: Bibliography), UCRL-6769, Office of Tech. Services, U.S.
Dept. of Commerce, Washington, 1962.
Lasers for Aerospace Weaponry , AF Aeronautical Systems Div., 1962.
(Available from Office of Technical Services, U.S. Dept. of Commerce,
Washington) .
Philip, N. A., Numerical Weather Prediction in Alt. , F. L. (Ed.) Advances
in Computers, Vol. I (see 125)
Read, T. , Command and Control , Center of Internat. Studies, Princeton
Univ., Policy memo 24, June 15, 1961.
Towes, C. (Ed.) Quantum Electronics , Colurabia Univ. Press, New York,
1960.
Vuylsteke, A., Elements of Maser Theory , Van Nostrand, Princeton,
1960.
5-48
5.3.1.2 Communications
Acklev; J. N.^ The Mul ti -Sequence Computer as a Communications Tool.
Proc. East. Joint Computer Conf., Dec. 1959.
Heckelman, T. J., and Lazinski^ R. H., Information-Handling in the
Defense Communications Control Complex. Proco East. Joint Computer
Conf.^ Deco 1961
Jackson, Wo (Edo), Communication Theory . Academic Press, New York, 1956.
Peterson, W. W., and Brown, D. T., Cyclic Codes for Error Detection .
Procc IRE, 49, 228-235, 1961.
Pierce, J, R., Symbols, Signals, and Noise . Harper, New York, 1961.
Saxby, E. P., Command Control , Economics Project - Communications
Progress Report . SDC, TM-875, Dec. 10, 1962.
Segal, R. J., and Guerber, H. P., Four Advanced Computers - Key to Air
Force Digital Data Communications System. Proc. East. Joint Computer
Conf., Dec. 1961.
Shannon, C. E., and Weaver, W., The Mathematical Theory of Communication
Univo of Illinois Press, Urbana, 1949.
Shaver, J. D., Tele-Processing Systems . Proco East. Joint Computer
Conf., Dec. 1961.
5.3.1.3 Language
Bar-Hillel, Y., The Present Status of Automatic Translation of Languages,
In Alto, F« Lo (Ed.), Advances in Computers, Vol. I (Sec. 125).
Chomsky, N., Syntactic Structures. Mouton, The Hague, 1957.
Foreign Developments in Machine Translation and Information Processing.
Uo So Joint Publications Research Service, JPRS 6633, Jan. 23, 1961
Hockett, C. F., A Course in Modern Linguistics . Macmillan, New York,
1958.
Klein, S., and Simmons, R. F., Syntactic Dependence and the Computer
Generation of Coherent Discourse . SDC, TM-758/000/00, Sept. 24, 1962.
5-49
Locke, W. N., and Booth, A. D. (Eds.), Machine Translation of Languages.
Wiley, New York, 1955; or M.loT. Press, Cambridge, 1955.
Shannon, C. Eo, Prediction and Entropy of Printed English. Bell System
Tecly. Journ., 30, 50-64, 1951.
Soviet Developments in Information Processing and Machine Translation.
U. S. Joint Publications Research Service, JPRS 3570, July 28, 1960.
Wiren, J., and Stubbs, H. L., Electronic Binary Selection for Phoneme
Classif ication o Journ. Acoust. Soc. Am, 28, 1082-1091, 1956,
Yngve, V. H., A Model and An Hypothesis for Language Structure. Proc
Am. Philosophical Soc, 104, 444-466, Oct. 1960.
} The Pepth Hypothesis in Structure pf L^ngu^gg ^nd
Its Mathematical Aspects. Am. Math. Society, 1961.
, Computer Programs for Translation. Scio Am., 206,
68-76, June 1962.
5.3.1.4 Displays, Consoles and Man-Machine Interaction
Green, R., et al., A Versatile Man-Machine Console. Proc. East. Joint
Computer Conf., Dec. 1961.
Kuehn, R. L., Dataview. A General -Purpose Data Display System. Proc.
Easto Joint Computer Conf., Dec. 1961.
Licklider, J. C. R., Man-Computer Symbiosis . Trans. IRE, HFE-9, 4-11,
1960.
J and Clark, W. E., On-Line Man-Computer Communication.
Proc. Spring Joint Computer Conf., May 1962o
Loewe, R. T., and Horowitz, P., Display System Design Considerations .
Proc. East. Joint Computer Conf«, Dec. 1961.
McCulloch, W. So (Ed.), Human Decisions in Complex Systems . Annals N. Y.
Acad, of Sci., 89, 715-896, 1961.
McRuer, D. T., and Krendel, E. S., Dynamic Responses of Human Operators.
USAF-WADC, TR 56-524, Oct. 1957.
5-50
Potts, To F., Ornstein, G. N., and Clymer, A. B., The Automatic
Determination of Human and Other System Parameters. Proc. West Joint
Computer Conf., May 1961.
Watson, Mo C, The Generation of Association Maps on a Digital Computer.
Publ ished Septo 17, 1962c
Wolin, Bo R., Are the Man and the Machine Relations? Proco Spring
Joint Computer Confo, May 1962.
5o3.1.5 Computer Technology
Alt., Fo Lo (Ed.), Advances in Computers . Vols. I and II, Academic Press,
New York; Vol. I, 1960, Vol. II, 1961.
Aoki, Mo, Estrin, G., and Tang, T., Parallelism in Computer Organization-
Random Number Generation in the Fixed-Plus-Variable Computer System.
Proco West. Joint Computer Conf., May 1961.
Bartee, T. Co, Digital Computer Fundamentals . McGraw-Hill, 1960
Blankenbaker, J. V., Logically Microprogrammed Computers . Trans. IRE,
EC-7, 103-109, 1958.
Bloom, L., Card Random Access Memory (CRAM): Functions and Use. Proc.
East. Joint Computer Conf., Dec. 1961.
Brown, G., et al.. Management and the Computer of the Future, MIT Press,
Cambridge, 1962.
Bradley, R. E., and Genna, J. F., Design of a One-Megacycle Iteration
Rate PDA . Proc. Spring Joint Computer Conf., May 1962.
Caldwell, S., Switching Circuits and Logical Design . Wiley, New York, 1958.
Campbell, E. K., The Determination of the Meanigful N-Tuples of
Instructions in a Computer Program. SDC, TM-865, Nov. 30, 1962.
Clapp, L. C, High-Speed Optical Computers and Quantum Transition
Memory Devices . Proc. West. Joint Computer Conf., May 1961.
Codd, E. F., et al.. Multiprogramming Stretch. Feasibility Considerations.
Comm. ACM, 2, 13-17, 1959.
, Multiprogram Scheduling. Comm. ACM, 3, 347-350 and 413-418,
1960
5-51
Coffman, E. G., The Qrgflni^;ation^] Pesign of Pigital Computers^ SDC,
FN-6881, Sept. 21, 1962.
Corbato, F. J., Merwin-Daggett, Mo, and Daley, R. C, An Experimental
Time-Sharing System . Proc. Spring Joint Computer Conf., May 1962.
Cox, Do R. and Smith, W. L., Queues . Methuen, London and Wiley, New
York.
Davies, So W., Design Objectives for the IBM Stretch Computer. Proc.
East. Joint Computer Conf., Dec. 1956.
Eckert, J. P., Uni vac-Larc, the Next Step In Computer Design. Proc.
East. Joint Computer Conf., Dec. 1956.
Eckman, D. P. (Ed.), Systems: Research and Design . (Proc. 1st Systems
Symposium at Case Inst, of Technology)
Franks, E., An Introduction to LUCID. SDC, FM-6837, August 28, 1962.
Gass, S. I., et al.. Project Mercury Real-Time Computational and
Data-Flow System: Part B - The Mercury Programming System. Proc. East.
Joint Computer Conf., Dec. 1961.
Gigacycle Computing Systems . AIEE Special Publication 5-136.
Gill, S., Parallel Programming. Cpmputer Journ., 1, 1-8, 1958.
Goldberg, J., and Green, M. W., Large Files for Information Retrieval
Based on Simultaneous Interrogation of All Items. In Proc. Symposium
on Large Capacity Memory Techniques for Computing Systems.
Heller, J., Sequencing Aspects of Multiprogramming. Journ. of ACM,
8, 426-439, 1961.
Hogan, D. L., Wigington, R. L., and Sears, R. W., Jr., Nanosecond
Computing . Internat. Sci. and Technology, Oct. 1962.
Holland, J., A Universal Cgmp^ter Capable of ^?^eciiting ^n Arbitrary
Number of Sub-Programs Simultaneously. Proc. East. Joint Computer
Conf., Deco 1959.
, iterative Circuit Computers. Proc. West. Joint Computer
Conf., May 1960.
5-52
Humphrey, W. S., Switching Circuits with Computer App lications. McGraw-
Hill, New York, 1958.
Kilburn, T., Payne, R. B., and Howarth, D. J., The Atlas Supervisor .
Proc. East. Joint Computer Confo, Dec 1961.
Kiseda, J. R., et al, A Magnetic Associative Memory. IBM Journ. of
Res. and Rev., 5, 106-121, 1961.
Leeds, Ho D., and Weinberg, G. M., Mu 1 1 i p r og r amm i ng . In Computer
Programming Fundamentalso McGraw-Hill, New York, 1961..
Maxwell, M. S., An Automatic Digital Data Assembly System for Space
Survei 1 lance . Proc. East. Joint Computer Conf., Dec. 1961.
McDermId, W. L., and Petersen, H. E., A Magnetic Associative Memory
System. IBM Journ. of Res. and Dev., 5, 59-62, 1961.
McGee, W. C, General i^^tlgn; Key to Successful Electronic Data
Processing . Journ. ACM, 6, 1-23, 1959.
Mealy, G. He, Operating Systems . Rand Rep. P-2584, 1962.
Miller, L., et al , A Mu]tl"LeYe] File Structure for information
Processing o Proc. West. Joint Computer Conf., May 1960.
Mittman, B., and Unger, A. (Eds.), Computer Aopl ications . 1960 .
MacMillan, New York, 1961.
Myers, P. B., A Survey of Microsystem Electronics. Proc. West. Joint
Computer Conf., May 1961.
Myhill, Je, Nerode, A., and Tennenbaum, S., Fundamental Concepts in the
Theory of Systems . USAF-WADC, Tech. Rep. No. 57-624, 1957.
Netherwood, D. B., Logical Machine Design: a Selected Bibliographyo
Transo IRE, EC-7, 155-178, 1958; and EC-8, 367-380, 1959.
Newell, A. (Ed.), Information-Processing Language V - Manual. Prentice-
Hall, New York, 1961.
Pfister, M., Jr«, Logical Design of Digital Computer. Wiley, New York,
1960.
5-53
Proceedings. Eastern Joint Computer Conference. Dec. 12-14. 1961,
(Vol. 20), Macmillan, New York, 1962.
Proceedings. Spring Joint Computer Conference, May 1-3. 1962. (Vol. 21),
National Press, Palo Alto, 1962.
Proceedings, Symposium on Large-Capacity Memory Techniques for Computing
Systems (Washington Do C. 1961 ). Macmillan, New York, 1962.
Prywes, N. S., and Gray, H. J., Jr., Multi-List Organized Associative
Memo ry o Moore School of Elect. Eng., Univ. of Pennsylvania, Jan. 1962.
Rachjmann, J. A., High Speed Computers . Proc. East. Joint Computer
Confo, Dec. 1959.
Rosin, R. F., An Organization of an Associative Cryogenic Computer.
Proc. Spring Joint Computer Conf., May 1962o
Rudd, D. F., Strategy of Data Selection . Op. Res., March-April, 1962.
Schoderbek, J. J., Some Weapon System Survival Probability Model s.
Op. Res., March-April, 1962.
Seeber, R. R., Jr., Cryogenic Associative Memory. Proc. Nat. Conf.
ACM, Aug. 23, 1960.
, Associative Self-Sorting Memory . Proc. East. Joint
Computer Conf., Dec. 1960.
., and Lindquist, A. B., Associative Memory with Ordered
Retrievel . IBM Journ. Res. and Dev., 6, 126-136, 1962/
Shafritz, A. B., Miller, A. E., and Rose, K., Multi -Level Programming
for a Real-Time System . Proc. East. Joint Computer Conf., Dec. 1961.
Shannon, c. E., The Synthesis of TwQ'Tgrniingl Switching Circuits* Bell
System Tech. Journ., 28, 59-98, 1949.
Shaw, J. C, et al., A Command Structure for Complex Information
Processing . Proc. West. Joint Computer Conf., May 1958.
Shoulders, K. R., Microelectronics Using Electron-Beam-Activated
Machining Techniques. In Alt, F. L., (Ed.), Advances in Computers,
Vol. II.
5-54
Strachey, C, Time Sharing in Large Fast Computers. In Proc. Internat.
Conf. on Information Processing, UNESCO.
Teager, Ho M., Real-Time Time-Shared Computer Project. Comm. ACM, 5
Japo 1962 - Research Summaries, 62.
von Bertalanffy, L., An Outline of General System Theory o Brit Journ.
Philc Sci., 1, 134-165, 1950.
West, G. P., Logical Organization of Computing Systems. SDC, SP-365,
June 15, 1961.
5c3.1.6 Theory of Automata
Burks, Ao W., and Wang, H., The Logic of Automata . Journ. Assoc.
Computing Macho, 4, 193-218, 279-297, 1947.
Burks, A. W., Computation, Behavior, and Structure in Fixed and Growing
Automata . In Yovits, M. C, and Cameron, S. (Ed.) Self-Organizing
Systems.
Chapuis, A., and Droz, E., Automata. A Historical and Technol igical
Study c Central Book, New York, 1958.
Church, Ao, Introduction to Mathematical Logic. Princeton Univo Press,
Princeton, 1956.
Copi, I. Mo, Symbo 1 i c Log i c . Macmillan, New York, 1954.
Davis, Mo, Computabi 1 i ty and Unsol vabi 1 i tv . McGraw-Hill, New York, 1958.
de Leeuw, Ko, et al, Computabi 1 ity by Probabilistic Machines. In Shannon,
Co E., and McCarthy, (Eds.) Automata Studies, 1956.
Edwards, W., Dynamic Decision Theory and Probabilistic Information
Processing. Human Factors, 4, No. 2, 1962.
Holland, J., A Survey of Automata Theory. Univ. of Michigan, 1959
(a Project Michigan memo).
Kleche, S. C, Introduction to Metamathematics. van Nostrand,
Princeton, 1952.
McNaughton, R., The Theory of Auto mata. A Survey. In Alt, F. L. (Ed.),
Advances in Computers, Vol. II.
5-55
Rabin, M. 0., and Scott, D., Finite Automata and Their Decision Problems.
IBM Journ. of Res. and Dev., 3, 114-125, 1959.
Shannon, C. E., Computers and Automata. Proc. Ire, 41, 1234-1241, 1953.
, and McCarthy, (Eds.), Automata Studies . Princeton Univ.
Press, Princeton, 1956.
Turing, A. M., On Computable Numbers with an Application to the
Entschei dunasproblem. Proc. London Math. Soc, 42, 230-265 (1936), and
43, 544-546 (1937).
von Foerster, H., Communication Amongst Automata . Amer. Journ.
Psychiatry, 118, 856-871, 1962.
von Neumann, Jo (Edo by Burks, A» W.), The Theory of Automata :
Construction, Reproduction and Homogeneity. Univ. of Illinois Press,
Urbana, 1962.
Yamada, Ho, A Mode of Real-Time Operations of a Subclass of Turing
Machines and the Existence of a Subclass of Recursive Functions which
are Not Real-Time Computable. Trans. IRE, EC-10, 1961.
5o3.1.7 Simulation Languages
1) CLP
Conway, R. W., Maxwell, W. L., and Walker, R. J., An Instruction Manual
for CORC - The Cornell Computing Language. Cornell University, Ithaca,
N. Yc, 1963.
Maxwell, Wo L., and Conway, R. W., CLP Preliminary Manual, Dept. of
Industrial Engineering . Cornell University, Ithaca, N. Y., No. 3, 9580,
October 1963, 22 ppo
Walker, W. E., and Delfausse, J. J., The Cornell List Processor , Ithaca,
N. Y., 1964.
2) CLS
Buxton, J. Eo, and Laski, J. G., "Control and Simulation Language, "Esso
Petroleum Co., Ltd., and IBM United Kingdom, Ltdo, London, England, August
1962, reprinted in the Computer Journal, Vol. 5, No. 3, 6 pp.
5-56
IBM United Kingdom, Ltd. and Esso Petroleum Co., Ltd., Control and
Simulation Language , Introductory Manual, March 1963, 39 pp.
IBM United Kingdom, Ltd. and Esso Petroleum Co., Ltd., Control and
Simulation Language , Reference Manual, March 1963, 95 pp.
3) DYNAMO
Pugh, Alexander L. , III, DYNAMO User's Manual . MIT Press, Cambridge,
Mass. , 1961.
4) GASP
U. S. Steel Company, GASP, a General Activity Simulation Program ,
Project No. 90.17-019(2), 1963, 52 pp.
5)
GPSS
Gordon, G. , A General Purpose Systems Simulator . IBM Systems Journal,
Vol. 1, September 1962, pp. 18-32.
Gordon, G. , A General Purpose Systems Simulator Program , Proc» EJCC.
MacMillan, New York, pp. 87-104.
IBM, Reference Manual, General Purpose Systems Simulator II , 1963, 149 pp.
6) SIMPAC
Lackner, M.R. , Toward A General Simulation Capability . Proc. of Western
Joint Computer Conference, 1962.
Systems Development Corp., SIMPAC User's Manual . Santa Monica, Calif.,
1962, TM-602/000/00.
7) SIMSCRIPT
Markowitz, H. , et al., S IMSCRIPT : A Simulation Programming Language .
RAND Memorandum RM-3310-PR, The Rand Corp., Santa Monica, Calif., 1962.
Prentice Hall, Englewood Cliff, N.J.
5-57
8) SOL
McNeley, John L. and Knuth, Donald E. , SOL - A Symbol ic Language for Gen -
eral-Purpose Systems Simulation , 1963, 45 pp.
9) OTHER
Kelley, D. H. , and Buxton, J. N. , Montecode - An interpretive Program
for Monte Carlo Simulations , Computer Journal, Jly 1962, pp. 88-93.
Ledley, R. S. , and Rotolo, L. S. , A Heuristic Concept and an Automatic
Computer Program Aid for Operational Simulation , Naval Research Logis-
tics (Quarterly, Vol. 9, 1962, pp. 231-244.
Tocher, K. D. , Handbook of the General Simulation Program , Vol. 1
(revise^) and Vol. II, The United Steel Companies Ltd., Sheffield,
England, Department of Operational Research and Cybernetics Report
77/ORC 3/ Tech. and Report 88/ORC 3 Tech.
Tocher, K. D. , and Owen, D.G., The Automatic Programming of Simulators ,
Proc. Second International Conference on Operational Research, English
Universities Press, 1960, p. 50.
5-58
5.3.2 Simulation
Adams, H. W., Generalized Modeling of Complex Systems . Seminar on
Simulation of Decision Systems at Mitre Corp., June 6, 7, 8, 1961.
Adams, R. H. and Jenkins, J. L., Simulation of Air Operations with the
Ai r-Battle Model , Operations Research, 8 Sept. - Oct. I960, p.600.
Alexander, Lawrence T., Man-Machine Simulation as a System Design and
Training Instrument . System Development Corp., SP-33 1/000/01 , Sept. 27,
1961 c
Arnold, C. R., Digital Simulation of a Conformal DIMUS Sonar System .
Phase 1, AD-265398, 28 Feb. 1961, p. 37-
Ashley, J. Robert, On the Analog Simulation of Mechanical Systems with
Stiff Position Limit Stops . Simulation, May 1964, p. 21.
Astronautics and Aeronautics, Control System Optimization Attained in
Record Time with Hybrid Simulation , June 1964, p. 7*
Bauer, W. F., Aspects of Real-Time Simulation . Symposium on Computers
in Simulation, Data Processing and Control, March 21, 1957"
Bekey, George A., Optimization of Mu 1 t i -Parameter Systems by Hybrid
Computer Technigues . Part 1, Simulation, Feb. 1964, p. 19.
Bekey, George A., Optimization of Mul t i -Parameter Systems by Hybrid
Computer Technigues , Part !l, Simulation, March 1964, p. 21.
Bishop, W» A. and Skillman, W- A., Digital Simulation of Pulse Doppler
Track-While-Scan Radar . IRE Internat. Convention Record., Vol. 10, Pt . 4,
p. 94.
Brotman, L., and Seid, B . , Digital Simulation of a Massed-Bomber, Manned-
interceptor Encounter ; Operations Research, 8 May - June I960, p. 421.
Cohen, 1. K., The Design and Objectives of Laboratory Problem IV . RM-3354-PR,
Rand Corp., Jan. 1963.
Conway, R. W., Johnson, B. M., Maxwell, W. L., A Queue Network Simulator ^
for the IBM 65O and Burroughs 220 . Communication of the ACM, Dec. 1969,
Vol . 2, p. 20.
5-59
Daev, D. S., Serdinov, A. I., Tarkhov, A. G., Model Simulation of Problems
Bearing Upon the Method of Radiowave Sound inq « I zv Akad Nauk USSR, Ser.
Geotiz, 1963, No. 6, p. 936 or (English trans.) Bull. Acad, Sci. USSR,
Geophys., Ser. No. 6 (June 1963, publ . Oct. 1963), p. 573.
EJCC, An Analog-Digital Simulator for the Design and Improvement of Man-
Machine Systems . 1957, Vol. 12, p. 90.
EJCC, Facilities and Instrumentation Required for Real-Time Simulation
involving System Hardware . 1957, Vol. 12, p. 96.
Elsinger, Robert C, Elimination of Steering Rate Gyro Bias in a Missile
Autop i lot . Simulation, Jan. 1964, p. 5.
Gainen, Leon, A Simulation Model for Data Systems Analysis , EJCC, 1961,
Vol . 20, p. 96.
Gordon, G., A General Purpose Systems Simulation Program . EJCC, I96I,
Vol . 20, p. 87.
Halbert, Peter W., Hybrid Simulation of an Aircraft Adaptive Control
System . WJCC, 1963, p. ^25-435.
Hamming, R. W., Computers and Simulation in Systems Engineering . IRE
Trans. Edue.,June 1962, Vol. E-5, No. 2, p. 76.
Hara, Hiroshi H., Special Technigues for Two-D imens ional Air-to-Air
Mi ssi le Simulation . May 1964, p. 29.
Hicks, C. L., Analog Simulation of an Acguisition and Tracking Radar
System with Command Capability . Eastern Simulation Councils, Mtg. 16
July 1962.
IBM B20-6346, General Purpose Systems Simulator II . Reference Manual.
Informatics Inc., RTDHS Primary Site Programming System .
Israel, David, Simulation Technigues for the Test and Evaluation of
Real-Time Computer Programs . ACM Journal, 1957, Vol. 4, p. 354.
Katz, J. H., Optimizing Bit-Time Computer Simulation . (TRW), Commun.
ACM 6, Nov. 1963, p. 679.
5-60
Kepcke, J., Computer Simulation of a Complex Secure Communications
System . Eastern Simulation Councils Mtg. 16 July 1962.
McLeod, John, Manned Spacecraft Simulation . Simulation, Fall 1963, p» 7»
Meissinger, He Fo, Simulation of Infrared Systems . Simulation, March 1964,
p. R-23c
Moore, C. J. and Lewis, T. So, Digital Simulation of Discrete Flow
Systems . Communications of the ACM, 3 December I960, pp. 659, 660, 662,
Redgrave, Michael J., Some approaches to Simulation. Modeling and
Gaining at SDC . System Development Corporation, March 19, 1962, SP-721.
Robin, F. A., Pardee, R. S., Scheffler, D. L. and Holland, F. C, A
Computer Driven Simulation Environment for Air Traffic Control Studies ,
WJCC, 1963, Vol . 24, p. 437-
Ross, A. He, Transmission Engineering by Computer Simulation . IRE Trans.
Commun. System, December 1962, Vol. CS-IO, No, 4, p. 457.
Smith, William E., A Digital Systems Simulator. WJCC . 1957, Vol. II,
p. 031.
System Development Corp., S imulat ion . BRT-12.
Thomas, 0. F., Analog-Digital Hybrid Computers in Simulation with
Humans and Hardware . WJCC, 1961, Vol. 19, p. 639.
Wiley, John S- Sons, Handbook of Automation Computation and Control .
Grabbe, Ramo Wooldridge, 1961.
WJCC, A Digital System Simulator . 1957, Vol. II, p. 31
WJCC, Real-Time Automobile Ride Simulation . I960, Vol. 17, p. 285.
WJCC, Simulation of an Information Channel on the IBM 704 Computer .
1959, Vol. 15, p. 87.
Wyatt, J. K., Prediction by Computer, Data Processing , 1 July -
September 1959, p. 137-